[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M69905NSBNX4V7BG16FJC2CW":3,"skill-same:01M69905NSBNX4V7BG16FJC2CW":91},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":23,"publishedAt":45,"linkOnly":46,"markdownContent":47,"markdownBody":48,"frontMatter":49,"refName":53,"commitSha":54,"skillDirectory":55,"forksCount":56,"repositoryHtmlUrl":57,"repositoryHomepageUrl":58,"ownerHtmlUrl":59,"ownerBlogUrl":60,"ownerDescription":61,"updatedAt":62,"install":63,"resources":76},"01M69905NSBNX4V7BG16FJC2CW","extract-python-environment-variables",null,"Scans a Python recipe to find every place an environment variable is accessed — including `os.environ.setdefault(\"V\", \"d\")`, whose \"d\" would otherwise be a hidden default a user editing .env.example has no way to discover — then ensures all variables are declared in `.env.example`, that `load_dotenv()` is bootstrapped in the package `__init__.py`, and that `python-dotenv>=1.0.0` is listed in `pyproject.toml`. When a new entry is added to `.env.example`, the extracted default from source is written as the value (with a `# extracted-by:extract-env-vars` marker and provenance comment); values that look like stubs (`\"my-project-id\"`, `\"changeme\"`, `\"\u003C...>\"`) are downgraded to the TODO placeholder but the source string is preserved in the marker comment. Also detects hardcoded model-name string literals (e.g. `\"gemini-3.5-flash\"` in `agent.py`) and rewrites them to an `os.getenv(...)` call. The variable name is derived from the assignment target when it names a model (`DEFAULT_EMBEDDING_MODEL` → `EMBEDDING_MODEL`), else `MODEL_NAME` (single model) or `MODEL_NAME_GENERATED_1` \u002F `MODEL_NAME_GENERATED_2`, … (multiple models). Normally no fallback default is written into the Python source; the one exception is when no `load_dotenv()` bootstrap could be installed (no package `__init__.py`), where the original literal is kept as the fallback so the lookup cannot evaluate to `None` at runtime. The model string is written as the value in `.env.example` with a comment prompting a rename.   When re-run against a recipe whose `.env.example` already has entries, the writer classifies each entry (skill-authored vs. user-authored, TODO vs. real value) and only rewrites lines it can prove it authored; user-authored lines are always preserved. A stale TODO (either skill-authored or a bare v1-era `\u003CTODO: update-this-value>` with no inline comment) is upgraded in place when source can supply a real default. IMPORTANT — two hard rules the skill NEVER breaks: (1) USER-EDIT SAFETY. Any `.env.example` line the skill cannot prove it authored is USER_OWNED and is never modified. The rewriter fails closed on any structural ambiguity (quoted values, backslash continuation, duplicate declarations) — a stale TODO left in place is cheap; a clobbered user edit is not. (2) ADDITIVE-ONLY FOR PYTHON FILES. The skill never writes new `os.environ.setdefault(...)` bootstrap lines into any Python file. Pre-existing `os.environ.setdefault(...)` or `os.getenv(\"VAR\", \"default\")` calls that the recipe author wrote by hand are LEFT UNTOUCHED — the skill's only writes to Python files are (a) the `load_dotenv()` bootstrap; (b) `# noqa: E402` on trailing relative imports that would otherwise trip Ruff; (c) hardcoded model-literal replacement. Use when the user wants to \"extract env vars\", \"update .env.example\", \"add load_dotenv\", \"surface setdefault defaults\", \"upgrade stale TODOs in .env.example\", \"replace hardcoded model names\", or \"fix environment variables\" in a Python recipe.","google\u002Fadk-recipes","google","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1342004?v=4","01M698VYEFDM0VJBZDYYJ2T83S","Google","01M6990124CWS5627VKGSKJXRD",10626,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fextract-python-environment-variables\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,9,39,40,41,42,43,44],"python","extract python environment variables","Variables","extract","environment variables","environment","todo","inline","at","re","dotenv","project","upgrade","writer","either","files","fix","update","run","placeholder","gemini","2026-11-10 10:12:30",false,"---\nname: extract-python-environment-variables\ndescription: >\n  Scans a Python recipe to find every place an environment variable is\n  accessed — including `os.environ.setdefault(\"V\", \"d\")`, whose \"d\" would\n  otherwise be a hidden default a user editing .env.example has no way to\n  discover — then ensures all variables are declared in `.env.example`,\n  that `load_dotenv()` is bootstrapped in the package `__init__.py`, and\n  that `python-dotenv>=1.0.0` is listed in `pyproject.toml`. When a new\n  entry is added to `.env.example`, the extracted default from source is\n  written as the value (with a `# extracted-by:extract-env-vars` marker\n  and provenance comment); values that look like stubs\n  (`\"my-project-id\"`, `\"changeme\"`, `\"\u003C...>\"`) are downgraded to the TODO\n  placeholder but the source string is preserved in the marker comment.\n  Also detects hardcoded model-name string literals (e.g.\n  `\"gemini-3.5-flash\"` in `agent.py`) and rewrites them to an\n  `os.getenv(...)` call. The variable name is derived from the assignment\n  target when it names a model (`DEFAULT_EMBEDDING_MODEL` →\n  `EMBEDDING_MODEL`), else `MODEL_NAME` (single model) or\n  `MODEL_NAME_GENERATED_1` \u002F `MODEL_NAME_GENERATED_2`, … (multiple models).\n  Normally no fallback default is written into the Python source; the one\n  exception is when no `load_dotenv()` bootstrap could be installed (no\n  package `__init__.py`), where the original literal is kept as the\n  fallback so the lookup cannot evaluate to `None` at runtime. The\n  model string is written as the value in `.env.example` with a comment\n  prompting a rename.   When re-run against a recipe whose `.env.example` already has entries,\n  the writer classifies each entry (skill-authored vs. user-authored,\n  TODO vs. real value) and only rewrites lines it can prove it authored;\n  user-authored lines are always preserved. A stale TODO (either\n  skill-authored or a bare v1-era `\u003CTODO: update-this-value>` with no\n  inline comment) is upgraded in place when source can supply a real\n  default.\n  IMPORTANT — two hard rules the skill NEVER breaks:\n  (1) USER-EDIT SAFETY. Any `.env.example` line the skill cannot prove\n  it authored is USER_OWNED and is never modified. The rewriter fails\n  closed on any structural ambiguity (quoted values, backslash\n  continuation, duplicate declarations) — a stale TODO left in place is\n  cheap; a clobbered user edit is not.\n  (2) ADDITIVE-ONLY FOR PYTHON FILES. The skill never writes new\n  `os.environ.setdefault(...)` bootstrap lines into any Python file.\n  Pre-existing `os.environ.setdefault(...)` or\n  `os.getenv(\"VAR\", \"default\")` calls that the recipe author wrote by\n  hand are LEFT UNTOUCHED — the skill's only writes to Python files are\n  (a) the `load_dotenv()` bootstrap; (b) `# noqa: E402` on trailing\n  relative imports that would otherwise trip Ruff; (c) hardcoded\n  model-literal replacement.\n  Use when the user wants to \"extract env vars\", \"update .env.example\",\n  \"add load_dotenv\", \"surface setdefault defaults\", \"upgrade stale TODOs\n  in .env.example\", \"replace hardcoded model names\", or \"fix environment\n  variables\" in a Python recipe.\nmetadata:\n  author: Google\n  license: Apache-2.0\n  version: 2.3.0\n---\n\n# Extract Python Environment Variables\n\nUse this skill to ensure a Python recipe properly declares and loads all\nenvironment variables it uses.\n\n---\n\n## What This Skill Does\n\nRuns `scripts\u002Fextract_env_vars.py` against a recipe directory. The script:\n\n1. **Scans** all `.py` files (excluding `tests\u002F` and common cache\u002Fvenv\n   directories) for environment variable accesses:\n   - `os.environ[\"VAR\"]`\n   - `os.environ.get(\"VAR\")` \u002F `os.environ.get(\"VAR\", \"default\")`\n   - `os.getenv(\"VAR\")` \u002F `os.getenv(\"VAR\", \"default\")`\n   - `os.environ.setdefault(\"VAR\", \"default\")` &nbsp;&nbsp;*(v2)*\n\n   Only names matching `^[A-Z_][A-Z0-9_]*$` (UPPER_SNAKE_CASE) are captured;\n   a lowercase name like `os.getenv(\"my_api_key\")` is **skipped** and a\n   `[WARN]` line lists any that were dropped. Rename such vars to uppercase\n   in source, or add them to `.env.example` by hand.\n\n2. **Updates `.env.example`** — appends any variables not already\n   declared AND upgrades stale TODO entries in place. Existing\n   user-authored lines are always preserved (see Rule 1 below).\n   - Creates `.env.example` from scratch if it does not exist.\n   - Each new\u002Fupgraded line has the shape:\n     ```\n     VAR=\u003Cvalue>  # extracted-by:extract-env-vars; \u003Cprovenance>\n     ```\n   - `\u003Cvalue>` is the **resolved default** extracted from source when the\n     author committed to one — with `os.environ.setdefault(...)` beating\n     `os.getenv(\"V\", \"d\")` \u002F `os.environ.get(\"V\", \"d\")` (setdefault is a\n     stronger commitment: it mutates the process environment, while getenv\n     fallback is per-read). Alphabetical file order breaks ties.\n   - Values that look like **stubs** are downgraded to\n     `\u003CTODO: update-this-value>` — e.g. `my-project-id`, `your-api-key`,\n     `changeme`, `\u003C...>`, anything containing `example.com`. The source\n     string is preserved in the marker comment so the maintainer sees\n     what was found and can fix the source too.\n   - When no source supplied a string-literal default, the value is\n     `\u003CTODO: update-this-value>` with a `no default in source` note.\n   - **Stale TODO upgrade (v2.1):** if an existing entry is either a\n     skill-authored TODO (has the marker) or a bare v1-era TODO (exact\n     value `\u003CTODO: update-this-value>`, no marker, no inline comment)\n     AND source can now supply a real default, the line is rewritten in\n     place. This closes the discoverability loop for recipes originally\n     processed by v1 (which always wrote TODOs regardless of what source\n     said).\n\n3. **Injects `load_dotenv()`** into the package `__init__.py` (the first\n   subdirectory inside the recipe that contains an `__init__.py`, skipping\n   `tests\u002F` and hidden directories). The snippet injected is:\n\n   ```python\n   from dotenv import load_dotenv\n\n   # Load variables from .env if present. In production the environment is\n   # already populated by the platform (Cloud Run, GKE, etc.), so a missing\n   # .env is expected and not an error.\n   load_dotenv()\n   ```\n\n   If `load_dotenv` is already present the injection is skipped.\n\n   **If no package `__init__.py` exists** (common in vertical plugins under\n   `plugins\u002F`, where the code lives in a plain `scripts\u002F` directory rather\n   than an importable package) the injection is skipped with a `[WARN]`,\n   and the step reports that no bootstrap is in place. Step 4 below depends\n   on that answer.\n\n   **Additionally — always, regardless of whether we injected** — appends\n   `# noqa: E402 -- must come after load_dotenv()` to any top-level relative\n   import (`from .x import y`) that sits AFTER a non-import module-level\n   statement. Two cases this covers:\n\n   - **Fresh injection.** The injected `load_dotenv()` call pushes\n     pre-existing trailing relative imports below a non-import statement, so\n     they'd trigger Ruff `E402` (\"module-level import not at top of file\")\n     when Phase 4 (ruff) of `prepare-python-recipe` runs.\n\n   - **Author-written bootstrap.** The recipe author already wrote\n     `load_dotenv()` + `os.environ.setdefault(...)` calls followed by a\n     trailing `from . import agent`, but never marked the trailing import.\n     The skill did NOT inject anything (load_dotenv was already present) but\n     still adds the noqa suffix so the file is lint-clean on the pipeline's\n     next ruff pass.\n\n   The suppression pass is precise — a relative import at the very TOP of\n   the file (before any non-import statement) is fine and left untouched.\n   Idempotent: a line that already carries `# noqa: E402` is skipped.\n\n4. **Replaces hardcoded model names** in source (e.g. `model=\"gemini-3.5-flash\"`\n   in `agent.py`) with an `os.getenv(...)` call.\n\n   **Position decides eligibility.** A model string is only promoted when it\n   is a *configurable constant*. Three positions mean it is something else,\n   and are left untouched:\n\n   - **Collection-literal entries** (dict keys and values, list\u002Fset\u002Ftuple\n     items) — lookup tables and enumerations of supported models:\n\n     ```python\n     IMAGE_MODELS = {\n         \"flash\": \"gemini-2.5-flash-image\",\n         \"pro\":   \"gemini-2.5-pro-image\",\n     }\n     ```\n\n     Rewriting the values collapses the table onto one env var; rewriting the\n     **keys** is worse, because `IMAGE_MODELS.get(\"flash\")` then never\n     matches anything. A dict key must stay a static literal.\n   - **Subscript indices** — `IMAGE_MODELS[\"gemini-3.1\"]` is a key *into*\n     such a table, so replacing it looks up a different entry.\n   - **Comparison operands** — `if \"gemini-3.1\" in model_id` tests a value\n     rather than configuring one.\n\n   Skipping a legitimate extraction is cheap (lift it by hand); silently\n   breaking a lookup table is not, so this errs towards skipping. Every\n   skipped literal is listed in an `[INFO]` block, so nothing is hidden — and\n   because these strings never reach the naming step, a dict target like\n   `IMAGE_MODELS` can no longer leak into `.env.example` as a variable name.\n\n   **Variable name.** The assignment target that holds the literal is used\n   when it names a model, since it carries far more meaning than a generic\n   fallback — `DEFAULT_EMBEDDING_MODEL = \"gemini-embedding-001\"` and\n   `embedding_model = cfg.get(\"embedding_model\", \"gemini-embedding-001\")`\n   both yield `EMBEDDING_MODEL`. This matters in recipes that already read a\n   *different* model var (say `GEMINI_MODEL` for the LLM), where a second\n   bare `MODEL_NAME` would be actively misleading. When a string is assigned\n   to conflicting target names, the most frequent wins, ties broken\n   alphabetically. Otherwise:\n   - A single unnamed model → `MODEL_NAME`\n   - The rest → `MODEL_NAME_GENERATED_1`, `MODEL_NAME_GENERATED_2`, … (sorted\n     alphabetically for determinism)\n\n   **Fallback argument.** Normally the emitted call is **bare** —\n   `os.getenv(\"EMBEDDING_MODEL\")`, no default — because default values are\n   the maintainer's decision, not the skill's. That is only safe when the\n   `load_dotenv()` bootstrap from step 3 is in place to populate the\n   environment.\n\n   **When step 3 could not install a bootstrap** (no package `__init__.py`),\n   nothing reads `.env`, so a bare lookup would evaluate to `None` at runtime\n   and silently break the recipe. In that case the original literal is\n   preserved as the fallback — `os.getenv(\"EMBEDDING_MODEL\",\n   \"gemini-embedding-001\")` — which keeps behaviour identical to before the\n   rewrite while still lifting the value into the environment. The step logs\n   an `[INFO]` line explaining the choice.\n\n   The actual model string is written as the value in `.env.example` (e.g.\n   `EMBEDDING_MODEL=gemini-embedding-001`) with a comment prompting the\n   maintainer to rename the variable if the derived name isn't right.\n\n5. **Updates `pyproject.toml`** — adds `python-dotenv>=1.0.0` to `[project]`\n   dependencies if it is not already there.\n\n---\n\n## Hard rules the skill NEVER breaks\n\n**Rule 1 — User-edit safety for `.env.example`.** The writer classifies\nevery existing entry before touching anything, and only rewrites lines\nit can prove it authored. Concretely:\n\n| Existing entry looks like | Source now provides | Action |\n|---|---|---|\n| Missing | any | **append** with resolved default (or TODO) |\n| A line the skill wrote (has `# extracted-by:extract-env-vars` marker) with `\u003CTODO: update-this-value>` value | a REAL default (not placeholder-shape) | **upgrade in place** |\n| Same as above | no default OR a placeholder-shape default | **skip** (would be TODO → TODO churn) |\n| Bare `VAR=\u003CTODO: update-this-value>` line with **no marker AND no inline comment** — a v1-era TODO | a REAL default | **upgrade in place** (v1 → v2 migration) |\n| A line the skill wrote with a REAL value | source has any value (even different) | **skip** (never overwrite skill-authored values on drift) |\n| Anything else (user typed value, user-authored comment, non-standard formatting) | any | **skip — never touched** |\n\nGolden rule underneath all of this: **fail closed**. If the classifier\ncan't confidently prove a line was skill-authored or is a bare v1 TODO,\nit's `USER_OWNED` and untouchable. If the rewriter finds a line whose\nstructure it doesn't understand (quoted value, backslash continuation,\nduplicate declarations), it refuses and warns rather than guessing. And\nbefore any write, the assembled file is re-parsed to verify every\nplanned upgrade re-parses as a skill-authored real value — a bug in the\nrewriter cannot silently corrupt `.env.example`.\n\nTo force regeneration of a line the skill would otherwise skip, delete\nthe line and re-run.\n\n*Rationale for extracting defaults (v2) and upgrading stale TODOs\n(v2.1): a value hidden inside a Python file that a user must know about\nbut has no reason to look at is, in practice, undocumented.\n`.env.example` is the documented configuration surface. Surfacing what\nthe author already committed to in code — and updating a stale TODO\nonce source can supply a real default — makes the configuration\ndiscoverable without inventing anything new. The value in source WAS\nthe value; we're only lifting it into view.*\n\n**Rule 2 — Additive-only for Python files.** The skill never writes new\n`os.environ.setdefault(...)` bootstrap lines into any Python file.\nPre-existing `os.environ.setdefault(...)` or\n`os.getenv(\"VAR\", \"default\")` calls that a recipe author wrote by hand\nare LEFT UNTOUCHED. The skill's only writes to Python files are:\n- Adding the `from dotenv import load_dotenv` + `load_dotenv()` snippet\n  (once, only if not already present).\n- Appending `# noqa: E402` to trailing relative imports that would\n  otherwise trip Ruff after the env-bootstrap block.\n- Replacing hardcoded model literals with `os.getenv(...)` calls — bare\n  when a `load_dotenv()` bootstrap is in place, otherwise retaining the\n  original literal as the fallback (see step 4 above).\n\nNote that scanning `os.environ.setdefault(...)` and lifting its value\ninto `.env.example` (v2) does NOT violate Rule 2 — the skill READS from\nPython source (unchanged behaviour) and WRITES only to `.env.example`.\n\n---\n\n## Rules for the Agent\n\n1. **Always use the script** — never manually edit `.env.example`, `__init__.py`,\n   or `pyproject.toml` to perform these changes.\n2. **Ask for the recipe directory** if the user has not provided one. Do not\n   assume a path.\n3. **After the script succeeds**, remind the user to:\n   - Copy `.env.example` → `.env` and fill in real values before running locally.\n   - Run `uv sync` to pick up the `python-dotenv` dependency if it was newly added.\n\n---\n\n## Input\n\n| Field | Required | Description |\n|-------|----------|-------------|\n| Recipe directory | Yes | Path to the recipe root (e.g. `contrib\u002Fpython\u002Fmy-recipe`, `core\u002Fpython\u002Fmy-recipe`, or `plugins\u002Fretail\u002Fstore-ops`) |\n\nIf the user has not specified the recipe directory, ask for it before proceeding.\n\n---\n\n## Run\n\nRun it through `uv` so it always executes on a Python 3.11+ interpreter — the\nscript uses the stdlib `tomllib`, which only exists from 3.11 onward. A bare\n`python3` that resolves to 3.9\u002F3.10 fails with `ModuleNotFoundError: tomllib`.\nNo `--with` packages are needed (the script is stdlib-only).\n\n```bash\nuv run --no-project python3 \\\n  .agents\u002Fskills\u002Fextract-python-environment-variables\u002Fscripts\u002Fextract_env_vars.py \\\n  --recipe-dir \u003CRECIPE_DIR>\n```\n\n### Preview first (optional)\n\nAdd `--dry-run` to report exactly what *would* change without modifying any\nfiles. Nothing is written to `.env.example`, `__init__.py`, `pyproject.toml`, or\nany source file. Useful for inspecting a recipe before committing to the edits:\n\n```bash\nuv run --no-project python3 \\\n  .agents\u002Fskills\u002Fextract-python-environment-variables\u002Fscripts\u002Fextract_env_vars.py \\\n  --recipe-dir \u003CRECIPE_DIR> --dry-run\n```\n\nIn dry-run output, actions are prefixed with `[DRY-RUN]` and phrased as\n\"Would add\" \u002F \"Would inject\" \u002F \"Would replace\".\n\n---\n\n## Respond\n\nDo not show the script's raw stdout. Reformat its results into clear Markdown\ntables (variables to add, files to update, model replacements) so they are easy\nto read — this is especially important for `--dry-run` output. For each variable\nor model string, include a column with the source file where it was found\n(locate it in the recipe's Python source; ignore `.env` and `.env.example`).\n\nFor added variables, surface the **resolved default** and **its provenance**\nso the maintainer knows which value the skill chose and where it came from:\n\n| Variable | Value written to .env.example | Source |\n|----------|-------------------------------|--------|\n| `GOOGLE_CLOUD_PROJECT` | `\u003CTODO: update-this-value>` *(downgraded — source had `\"my-project-id\"`)* | `agent.py` (setdefault) |\n| `GOOGLE_GENAI_USE_VERTEXAI` | `TRUE` | `__init__.py` (setdefault) |\n| `LOG_LEVEL` | `INFO` | `logging_setup.py` (getenv fallback) |\n| `API_KEY` | `\u003CTODO: update-this-value>` *(no default in source)* | `client.py` (getenv, no fallback) |\n\nIf the script reports **upgrades** (stale TODO entries rewritten in\nplace with real defaults from source), surface those in a separate table\nso the maintainer can eyeball each one:\n\n| Variable | Was | Now | Source |\n|----------|-----|-----|--------|\n| `USE_VERTEX` | `\u003CTODO: update-this-value>` | `TRUE` | `__init__.py` (setdefault) |\n\nIf the script logged `[WARN] Refused to upgrade …` for any variable,\nsurface that too — the line's structure was ambiguous and was left\nuntouched deliberately.\n\nOnce the script finishes successfully, summarise what changed:\n\n- Which variables were **added** to `.env.example` (with resolved value +\n  source) or confirm it was already up to date.\n- Which existing TODOs were **upgraded** in place (v2.1) with real\n  defaults from source.\n- Any values that were **downgraded to `\u003CTODO>` because they looked like\n  placeholders** — the maintainer should fix the source too.\n- Whether `load_dotenv()` was injected or was already present.\n- Whether `python-dotenv` was added to `pyproject.toml` or was already there.\n\nThen remind the user of the next steps:\n\n```\nNext steps:\n  cp \u003CRECIPE_DIR>\u002F.env.example \u003CRECIPE_DIR>\u002F.env   # then fill in real values\n  cd \u003CRECIPE_DIR> && uv sync                        # install python-dotenv if newly added\n```\n\nDo not make any further changes. End your turn.\n","# Extract Python Environment Variables\n\nUse this skill to ensure a Python recipe properly declares and loads all\nenvironment variables it uses.\n\n---\n\n## What This Skill Does\n\nRuns `scripts\u002Fextract_env_vars.py` against a recipe directory. The script:\n\n1. **Scans** all `.py` files (excluding `tests\u002F` and common cache\u002Fvenv\n   directories) for environment variable accesses:\n   - `os.environ[\"VAR\"]`\n   - `os.environ.get(\"VAR\")` \u002F `os.environ.get(\"VAR\", \"default\")`\n   - `os.getenv(\"VAR\")` \u002F `os.getenv(\"VAR\", \"default\")`\n   - `os.environ.setdefault(\"VAR\", \"default\")` &nbsp;&nbsp;*(v2)*\n\n   Only names matching `^[A-Z_][A-Z0-9_]*$` (UPPER_SNAKE_CASE) are captured;\n   a lowercase name like `os.getenv(\"my_api_key\")` is **skipped** and a\n   `[WARN]` line lists any that were dropped. Rename such vars to uppercase\n   in source, or add them to `.env.example` by hand.\n\n2. **Updates `.env.example`** — appends any variables not already\n   declared AND upgrades stale TODO entries in place. Existing\n   user-authored lines are always preserved (see Rule 1 below).\n   - Creates `.env.example` from scratch if it does not exist.\n   - Each new\u002Fupgraded line has the shape:\n     ```\n     VAR=\u003Cvalue>  # extracted-by:extract-env-vars; \u003Cprovenance>\n     ```\n   - `\u003Cvalue>` is the **resolved default** extracted from source when the\n     author committed to one — with `os.environ.setdefault(...)` beating\n     `os.getenv(\"V\", \"d\")` \u002F `os.environ.get(\"V\", \"d\")` (setdefault is a\n     stronger commitment: it mutates the process environment, while getenv\n     fallback is per-read). Alphabetical file order breaks ties.\n   - Values that look like **stubs** are downgraded to\n     `\u003CTODO: update-this-value>` — e.g. `my-project-id`, `your-api-key`,\n     `changeme`, `\u003C...>`, anything containing `example.com`. The source\n     string is preserved in the marker comment so the maintainer sees\n     what was found and can fix the source too.\n   - When no source supplied a string-literal default, the value is\n     `\u003CTODO: update-this-value>` with a `no default in source` note.\n   - **Stale TODO upgrade (v2.1):** if an existing entry is either a\n     skill-authored TODO (has the marker) or a bare v1-era TODO (exact\n     value `\u003CTODO: update-this-value>`, no marker, no inline comment)\n     AND source can now supply a real default, the line is rewritten in\n     place. This closes the discoverability loop for recipes originally\n     processed by v1 (which always wrote TODOs regardless of what source\n     said).\n\n3. **Injects `load_dotenv()`** into the package `__init__.py` (the first\n   subdirectory inside the recipe that contains an `__init__.py`, skipping\n   `tests\u002F` and hidden directories). The snippet injected is:\n\n   ```python\n   from dotenv import load_dotenv\n\n   # Load variables from .env if present. In production the environment is\n   # already populated by the platform (Cloud Run, GKE, etc.), so a missing\n   # .env is expected and not an error.\n   load_dotenv()\n   ```\n\n   If `load_dotenv` is already present the injection is skipped.\n\n   **If no package `__init__.py` exists** (common in vertical plugins under\n   `plugins\u002F`, where the code lives in a plain `scripts\u002F` directory rather\n   than an importable package) the injection is skipped with a `[WARN]`,\n   and the step reports that no bootstrap is in place. Step 4 below depends\n   on that answer.\n\n   **Additionally — always, regardless of whether we injected** — appends\n   `# noqa: E402 -- must come after load_dotenv()` to any top-level relative\n   import (`from .x import y`) that sits AFTER a non-import module-level\n   statement. Two cases this covers:\n\n   - **Fresh injection.** The injected `load_dotenv()` call pushes\n     pre-existing trailing relative imports below a non-import statement, so\n     they'd trigger Ruff `E402` (\"module-level import not at top of file\")\n     when Phase 4 (ruff) of `prepare-python-recipe` runs.\n\n   - **Author-written bootstrap.** The recipe author already wrote\n     `load_dotenv()` + `os.environ.setdefault(...)` calls followed by a\n     trailing `from . import agent`, but never marked the trailing import.\n     The skill did NOT inject anything (load_dotenv was already present) but\n     still adds the noqa suffix so the file is lint-clean on the pipeline's\n     next ruff pass.\n\n   The suppression pass is precise — a relative import at the very TOP of\n   the file (before any non-import statement) is fine and left untouched.\n   Idempotent: a line that already carries `# noqa: E402` is skipped.\n\n4. **Replaces hardcoded model names** in source (e.g. `model=\"gemini-3.5-flash\"`\n   in `agent.py`) with an `os.getenv(...)` call.\n\n   **Position decides eligibility.** A model string is only promoted when it\n   is a *configurable constant*. Three positions mean it is something else,\n   and are left untouched:\n\n   - **Collection-literal entries** (dict keys and values, list\u002Fset\u002Ftuple\n     items) — lookup tables and enumerations of supported models:\n\n     ```python\n     IMAGE_MODELS = {\n         \"flash\": \"gemini-2.5-flash-image\",\n         \"pro\":   \"gemini-2.5-pro-image\",\n     }\n     ```\n\n     Rewriting the values collapses the table onto one env var; rewriting the\n     **keys** is worse, because `IMAGE_MODELS.get(\"flash\")` then never\n     matches anything. A dict key must stay a static literal.\n   - **Subscript indices** — `IMAGE_MODELS[\"gemini-3.1\"]` is a key *into*\n     such a table, so replacing it looks up a different entry.\n   - **Comparison operands** — `if \"gemini-3.1\" in model_id` tests a value\n     rather than configuring one.\n\n   Skipping a legitimate extraction is cheap (lift it by hand); silently\n   breaking a lookup table is not, so this errs towards skipping. Every\n   skipped literal is listed in an `[INFO]` block, so nothing is hidden — and\n   because these strings never reach the naming step, a dict target like\n   `IMAGE_MODELS` can no longer leak into `.env.example` as a variable name.\n\n   **Variable name.** The assignment target that holds the literal is used\n   when it names a model, since it carries far more meaning than a generic\n   fallback — `DEFAULT_EMBEDDING_MODEL = \"gemini-embedding-001\"` and\n   `embedding_model = cfg.get(\"embedding_model\", \"gemini-embedding-001\")`\n   both yield `EMBEDDING_MODEL`. This matters in recipes that already read a\n   *different* model var (say `GEMINI_MODEL` for the LLM), where a second\n   bare `MODEL_NAME` would be actively misleading. When a string is assigned\n   to conflicting target names, the most frequent wins, ties broken\n   alphabetically. Otherwise:\n   - A single unnamed model → `MODEL_NAME`\n   - The rest → `MODEL_NAME_GENERATED_1`, `MODEL_NAME_GENERATED_2`, … (sorted\n     alphabetically for determinism)\n\n   **Fallback argument.** Normally the emitted call is **bare** —\n   `os.getenv(\"EMBEDDING_MODEL\")`, no default — because default values are\n   the maintainer's decision, not the skill's. That is only safe when the\n   `load_dotenv()` bootstrap from step 3 is in place to populate the\n   environment.\n\n   **When step 3 could not install a bootstrap** (no package `__init__.py`),\n   nothing reads `.env`, so a bare lookup would evaluate to `None` at runtime\n   and silently break the recipe. In that case the original literal is\n   preserved as the fallback — `os.getenv(\"EMBEDDING_MODEL\",\n   \"gemini-embedding-001\")` — which keeps behaviour identical to before the\n   rewrite while still lifting the value into the environment. The step logs\n   an `[INFO]` line explaining the choice.\n\n   The actual model string is written as the value in `.env.example` (e.g.\n   `EMBEDDING_MODEL=gemini-embedding-001`) with a comment prompting the\n   maintainer to rename the variable if the derived name isn't right.\n\n5. **Updates `pyproject.toml`** — adds `python-dotenv>=1.0.0` to `[project]`\n   dependencies if it is not already there.\n\n---\n\n## Hard rules the skill NEVER breaks\n\n**Rule 1 — User-edit safety for `.env.example`.** The writer classifies\nevery existing entry before touching anything, and only rewrites lines\nit can prove it authored. Concretely:\n\n| Existing entry looks like | Source now provides | Action |\n|---|---|---|\n| Missing | any | **append** with resolved default (or TODO) |\n| A line the skill wrote (has `# extracted-by:extract-env-vars` marker) with `\u003CTODO: update-this-value>` value | a REAL default (not placeholder-shape) | **upgrade in place** |\n| Same as above | no default OR a placeholder-shape default | **skip** (would be TODO → TODO churn) |\n| Bare `VAR=\u003CTODO: update-this-value>` line with **no marker AND no inline comment** — a v1-era TODO | a REAL default | **upgrade in place** (v1 → v2 migration) |\n| A line the skill wrote with a REAL value | source has any value (even different) | **skip** (never overwrite skill-authored values on drift) |\n| Anything else (user typed value, user-authored comment, non-standard formatting) | any | **skip — never touched** |\n\nGolden rule underneath all of this: **fail closed**. If the classifier\ncan't confidently prove a line was skill-authored or is a bare v1 TODO,\nit's `USER_OWNED` and untouchable. If the rewriter finds a line whose\nstructure it doesn't understand (quoted value, backslash continuation,\nduplicate declarations), it refuses and warns rather than guessing. And\nbefore any write, the assembled file is re-parsed to verify every\nplanned upgrade re-parses as a skill-authored real value — a bug in the\nrewriter cannot silently corrupt `.env.example`.\n\nTo force regeneration of a line the skill would otherwise skip, delete\nthe line and re-run.\n\n*Rationale for extracting defaults (v2) and upgrading stale TODOs\n(v2.1): a value hidden inside a Python file that a user must know about\nbut has no reason to look at is, in practice, undocumented.\n`.env.example` is the documented configuration surface. Surfacing what\nthe author already committed to in code — and updating a stale TODO\nonce source can supply a real default — makes the configuration\ndiscoverable without inventing anything new. The value in source WAS\nthe value; we're only lifting it into view.*\n\n**Rule 2 — Additive-only for Python files.** The skill never writes new\n`os.environ.setdefault(...)` bootstrap lines into any Python file.\nPre-existing `os.environ.setdefault(...)` or\n`os.getenv(\"VAR\", \"default\")` calls that a recipe author wrote by hand\nare LEFT UNTOUCHED. The skill's only writes to Python files are:\n- Adding the `from dotenv import load_dotenv` + `load_dotenv()` snippet\n  (once, only if not already present).\n- Appending `# noqa: E402` to trailing relative imports that would\n  otherwise trip Ruff after the env-bootstrap block.\n- Replacing hardcoded model literals with `os.getenv(...)` calls — bare\n  when a `load_dotenv()` bootstrap is in place, otherwise retaining the\n  original literal as the fallback (see step 4 above).\n\nNote that scanning `os.environ.setdefault(...)` and lifting its value\ninto `.env.example` (v2) does NOT violate Rule 2 — the skill READS from\nPython source (unchanged behaviour) and WRITES only to `.env.example`.\n\n---\n\n## Rules for the Agent\n\n1. **Always use the script** — never manually edit `.env.example`, `__init__.py`,\n   or `pyproject.toml` to perform these changes.\n2. **Ask for the recipe directory** if the user has not provided one. Do not\n   assume a path.\n3. **After the script succeeds**, remind the user to:\n   - Copy `.env.example` → `.env` and fill in real values before running locally.\n   - Run `uv sync` to pick up the `python-dotenv` dependency if it was newly added.\n\n---\n\n## Input\n\n| Field | Required | Description |\n|-------|----------|-------------|\n| Recipe directory | Yes | Path to the recipe root (e.g. `contrib\u002Fpython\u002Fmy-recipe`, `core\u002Fpython\u002Fmy-recipe`, or `plugins\u002Fretail\u002Fstore-ops`) |\n\nIf the user has not specified the recipe directory, ask for it before proceeding.\n\n---\n\n## Run\n\nRun it through `uv` so it always executes on a Python 3.11+ interpreter — the\nscript uses the stdlib `tomllib`, which only exists from 3.11 onward. A bare\n`python3` that resolves to 3.9\u002F3.10 fails with `ModuleNotFoundError: tomllib`.\nNo `--with` packages are needed (the script is stdlib-only).\n\n```bash\nuv run --no-project python3 \\\n  .agents\u002Fskills\u002Fextract-python-environment-variables\u002Fscripts\u002Fextract_env_vars.py \\\n  --recipe-dir \u003CRECIPE_DIR>\n```\n\n### Preview first (optional)\n\nAdd `--dry-run` to report exactly what *would* change without modifying any\nfiles. Nothing is written to `.env.example`, `__init__.py`, `pyproject.toml`, or\nany source file. Useful for inspecting a recipe before committing to the edits:\n\n```bash\nuv run --no-project python3 \\\n  .agents\u002Fskills\u002Fextract-python-environment-variables\u002Fscripts\u002Fextract_env_vars.py \\\n  --recipe-dir \u003CRECIPE_DIR> --dry-run\n```\n\nIn dry-run output, actions are prefixed with `[DRY-RUN]` and phrased as\n\"Would add\" \u002F \"Would inject\" \u002F \"Would replace\".\n\n---\n\n## Respond\n\nDo not show the script's raw stdout. Reformat its results into clear Markdown\ntables (variables to add, files to update, model replacements) so they are easy\nto read — this is especially important for `--dry-run` output. For each variable\nor model string, include a column with the source file where it was found\n(locate it in the recipe's Python source; ignore `.env` and `.env.example`).\n\nFor added variables, surface the **resolved default** and **its provenance**\nso the maintainer knows which value the skill chose and where it came from:\n\n| Variable | Value written to .env.example | Source |\n|----------|-------------------------------|--------|\n| `GOOGLE_CLOUD_PROJECT` | `\u003CTODO: update-this-value>` *(downgraded — source had `\"my-project-id\"`)* | `agent.py` (setdefault) |\n| `GOOGLE_GENAI_USE_VERTEXAI` | `TRUE` | `__init__.py` (setdefault) |\n| `LOG_LEVEL` | `INFO` | `logging_setup.py` (getenv fallback) |\n| `API_KEY` | `\u003CTODO: update-this-value>` *(no default in source)* | `client.py` (getenv, no fallback) |\n\nIf the script reports **upgrades** (stale TODO entries rewritten in\nplace with real defaults from source), surface those in a separate table\nso the maintainer can eyeball each one:\n\n| Variable | Was | Now | Source |\n|----------|-----|-----|--------|\n| `USE_VERTEX` | `\u003CTODO: update-this-value>` | `TRUE` | `__init__.py` (setdefault) |\n\nIf the script logged `[WARN] Refused to upgrade …` for any variable,\nsurface that too — the line's structure was ambiguous and was left\nuntouched deliberately.\n\nOnce the script finishes successfully, summarise what changed:\n\n- Which variables were **added** to `.env.example` (with resolved value +\n  source) or confirm it was already up to date.\n- Which existing TODOs were **upgraded** in place (v2.1) with real\n  defaults from source.\n- Any values that were **downgraded to `\u003CTODO>` because they looked like\n  placeholders** — the maintainer should fix the source too.\n- Whether `load_dotenv()` was injected or was already present.\n- Whether `python-dotenv` was added to `pyproject.toml` or was already there.\n\nThen remind the user of the next steps:\n\n```\nNext steps:\n  cp \u003CRECIPE_DIR>\u002F.env.example \u003CRECIPE_DIR>\u002F.env   # then fill in real values\n  cd \u003CRECIPE_DIR> && uv sync                        # install python-dotenv if newly added\n```\n\nDo not make any further changes. End your turn.\n",{"name":5,"metadata":50,"description":52},{"author":12,"license":15,"version":51},"2.3.0","Scans a Python recipe to find every place an environment variable is accessed — including `os.environ.setdefault(\"V\", \"d\")`, whose \"d\" would otherwise be a hidden default a user editing .env.example has no way to discover — then ensures all variables are declared in `.env.example`, that `load_dotenv()` is bootstrapped in the package `__init__.py`, and that `python-dotenv>=1.0.0` is listed in `pyproject.toml`. When a new entry is added to `.env.example`, the extracted default from source is written as the value (with a `# extracted-by:extract-env-vars` marker and provenance comment); values that look like stubs (`\"my-project-id\"`, `\"changeme\"`, `\"\u003C...>\"`) are downgraded to the TODO placeholder but the source string is preserved in the marker comment. Also detects hardcoded model-name string literals (e.g. `\"gemini-3.5-flash\"` in `agent.py`) and rewrites them to an `os.getenv(...)` call. The variable name is derived from the assignment target when it names a model (`DEFAULT_EMBEDDING_MODEL` → `EMBEDDING_MODEL`), else `MODEL_NAME` (single model) or `MODEL_NAME_GENERATED_1` \u002F `MODEL_NAME_GENERATED_2`, … (multiple models). Normally no fallback default is written into the Python source; the one exception is when no `load_dotenv()` bootstrap could be installed (no package `__init__.py`), where the original literal is kept as the fallback so the lookup cannot evaluate to `None` at runtime. The model string is written as the value in `.env.example` with a comment prompting a rename.   When re-run against a recipe whose `.env.example` already has entries, the writer classifies each entry (skill-authored vs. user-authored, TODO vs. real value) and only rewrites lines it can prove it authored; user-authored lines are always preserved. A stale TODO (either skill-authored or a bare v1-era `\u003CTODO: update-this-value>` with no inline comment) is upgraded in place when source can supply a real default. IMPORTANT — two hard rules the skill NEVER breaks: (1) USER-EDIT SAFETY. Any `.env.example` line the skill cannot prove it authored is USER_OWNED and is never modified. The rewriter fails closed on any structural ambiguity (quoted values, backslash continuation, duplicate declarations) — a stale TODO left in place is cheap; a clobbered user edit is not. (2) ADDITIVE-ONLY FOR PYTHON FILES. The skill never writes new `os.environ.setdefault(...)` bootstrap lines into any Python file. Pre-existing `os.environ.setdefault(...)` or `os.getenv(\"VAR\", \"default\")` calls that the recipe author wrote by hand are LEFT UNTOUCHED — the skill's only writes to Python files are (a) the `load_dotenv()` bootstrap; (b) `# noqa: E402` on trailing relative imports that would otherwise trip Ruff; (c) hardcoded model-literal replacement. Use when the user wants to \"extract env vars\", \"update .env.example\", \"add load_dotenv\", \"surface setdefault defaults\", \"upgrade stale TODOs in .env.example\", \"replace hardcoded model names\", or \"fix environment variables\" in a Python recipe.\n","main","665f7b2a181686f2ba3c6c942e9b30a084b12ec3",".agents\u002Fskills\u002Fextract-python-environment-variables",2962,"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes","https:\u002F\u002Fadk.dev","https:\u002F\u002Fgithub.com\u002Fgoogle","https:\u002F\u002Fopensource.google\u002F","Google ❤️ Open Source","2026-11-14 04:01:32",{"source":8,"skillName":5,"commands":64,"prompt":74,"zipAvailable":75},[65,68,71],{"tool":66,"command":67},"npx","npx skills add google\u002Fadk-recipes --skill extract-python-environment-variables",{"tool":69,"command":70},"bunx","bunx skills add google\u002Fadk-recipes --skill extract-python-environment-variables",{"tool":72,"command":73},"pnpm","pnpm dlx skills add google\u002Fadk-recipes --skill extract-python-environment-variables","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fextract-python-environment-variables\u002FSKILL.md",true,[77,83,87],{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},"scripts\u002Fextract_env_vars.py","SCRIPT","脚本",102141,"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fextract-python-environment-variables\u002Fscripts\u002Fextract_env_vars.py",{"path":84,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":85,"sourceHtmlUrl":86},"tests\u002Fconftest.py",919,"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fextract-python-environment-variables\u002Ftests\u002Fconftest.py",{"path":88,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":89,"sourceHtmlUrl":90},"tests\u002Ftest_extract_env_vars.py",120324,"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fextract-python-environment-variables\u002Ftests\u002Ftest_extract_env_vars.py",[92,113,146,171,188,209],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":105,"publishedAt":45},"01M6991EKEG8XEM3VWW4VGV0ZQ","retail-virtual-tryon","Creates virtual try-on agents supporting image and video (catwalk animation) try-ons on Google Cloud (Gemini image models and Veo on Gemini Enterprise Agent Platform). Handles resource setup, user photo uploading, image\u002Fvideo generation pipelines, local testing, and evaluation.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002Fplugins\u002Fretail\u002Fskills\u002Fvirtual-tryon\u002FSKILL.md",[98,101,104],{"categoryNo":99,"categoryName":100,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":102,"categoryName":103,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000016","图像与多媒体",{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[106,107,108,9,109,44,110,111,112],"retail virtual tryon","google cloud","image","testing","genai","adk","gemini enterprise agent platform",{"skillNo":114,"name":115,"nameZh":6,"description":116,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":117,"categories":118,"keywords":128,"publishedAt":45},"01M6991ABTZW7FCPREXY8GD4RG","retail-product-search","Creates product search agents with semantic search and RAG on Google Cloud (Vector Search on Gemini Enterprise Agent Platform, BigQuery, embeddings). Use when the user wants to \"build a product search agent\", \"create an e-commerce search\", \"make a shopping assistant\", \"set up semantic catalog discovery\", \"ingest products into Vector Search\", or \"deploy a retail RAG agent\". Handles the full pipeline: catalog data ingestion to BigQuery, Gemini Enterprise Agent Platform Vector Search collection setup, ADK agent scaffolding, evaluation, and Cloud Run deployment.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002Fplugins\u002Fretail\u002Fskills\u002Fproduct-search\u002FSKILL.md",[119,122,123],{"categoryNo":120,"categoryName":121,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000011","运维与部署",{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":124,"categoryName":125,"taxonomyType":126,"taxonomyTypeDesc":127},"01JBSEEDCAT000000000000031","产品经理","PROFESSION","适用职业",[129,130,131,112,111,44,132,9,107,133,134,135,136,137,138,139,140,141,142,143,42,144,145],"retail product search","product","search","assistant","pipeline","when","commerce","scaffolding","shopping","up","ingestion","collection","SET","vector","apache","create","rag",{"skillNo":147,"name":148,"nameZh":6,"description":149,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":150,"categories":151,"keywords":153,"publishedAt":45},"01M69918HZCYZHD7K9Y9ZSD287","routines","Use for a recurring unattended task (cron; pull data, open a PR, post a digest), or to test\u002Fdry-run a routine first. Explains authoring and testing via the routine tool.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002Fcore\u002Fpython\u002Flong-horizon-harness\u002Fhorizon\u002Fbuiltin_skills\u002Froutines\u002FSKILL.md",[152],{"categoryNo":99,"categoryName":100,"taxonomyType":21,"taxonomyTypeDesc":22},[148,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,42,169,109,170],"dry","routine","for","post","via","authoring","digest","task","use","recurring","pr","tool","open","first","pull","test","data",{"skillNo":172,"name":173,"nameZh":6,"description":174,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":175,"categories":176,"keywords":177,"publishedAt":45},"01M699178ZDFDXEZ11DG1Y9CTF","policy","Inspect the per-workspace tool-policy overlay (.lha\u002Fpolicies.jsonl) that gates destructive or sensitive tool calls, and tell the user what to change if they want a rule added or removed.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002Fcore\u002Fpython\u002Flong-horizon-harness\u002Fhorizon\u002Fbuiltin_skills\u002Fpolicy\u002FSKILL.md",[],[173,178,179,180,181,182,183,184,185,186,165,187],"gates","overlay","tool calls","inspect","calls","jsonl","rule","user","workspace","change",{"skillNo":189,"name":190,"nameZh":6,"description":191,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":192,"categories":193,"keywords":194,"publishedAt":45},"01M69915F8591SQBA1ZHK6NDA7","google-workspace","Read\u002Fwrite Google Drive, Docs, Sheets, Gmail, Calendar, Chat, Tasks, Slides via the `gws` CLI. Needs an OAuth client or service account configured first. Use for any Google Workspace task.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002Fcore\u002Fpython\u002Flong-horizon-harness\u002Fhorizon\u002Fbuiltin_skills\u002Fgoogle-workspace\u002FSKILL.md",[],[186,195,9,196,162,197,198,199,167,200,201,202,203,204,161,156,205,206,158,207,208],"google workspace","calendar","write","read","slides","OAuth","tasks","docs","cli","gmail","sheets","client","chat","gws",{"skillNo":210,"name":211,"nameZh":6,"description":212,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":213,"categories":214,"keywords":216,"publishedAt":45},"01M69914416KTZ6Z0WD425YYJ0","find-skills","Discover\u002Finstall NEW agent skills from external sources (skills.sh). Use for capabilities you lack, e.g. \"find a skill for X\". Do NOT use to list skills you already have; see `\u003Cavailable_skills>`.","https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fadk-recipes\u002Fblob\u002Fmain\u002Fcore\u002Fpython\u002Flong-horizon-harness\u002Fhorizon\u002Fbuiltin_skills\u002Ffind-skills\u002FSKILL.md",[215],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[217,218,156,219,220,221,222,223,162,224,225,226,227],"find skills","find","you","sources","discover","new agent","do","install","list","list skills","capabilities"]