replace-me-skill

<!-- Authoring checklist (delete this comment block before committing):

1. Copy: cp -r templates/skill skills/<domain>/<my-new-skill>, then mv replace_me.py <my_new_skill>.py and rename the tests/ file. The folder name must equal the frontmatter name: load_skill finds skills by folder name.

2. Write the frontmatter above by hand. It is the ONLY metadata the agent sees, and omicsclaw/skills/ reads exactly four keys:

   name required, unique across all skills. A duplicate is skipped, not merged. This is what use_skill and load_skill are called with. description required. The one line the model routes on, and the only thing about this skill in the system prompt. Say when to LOAD it and when to SKIP it, naming the skill to use instead — the skip half is what prevents a wrong choice. trigger optional, comma-separated. Discovery keywords for /skills <query>. It does NOT auto-fire the skill. tags optional. Also only /skills <query>.

   Anything else is inert: it is neither read nor validated.

3. Put the computations in _api.py (rules in templates/skill/README.md), then regenerate the ## API section: python skills/_sdk/notebook/run.py api skills/<domain>/<my-new-skill> --write A test compares it with _api.py on every run.

4. Write the other sections and the references/*.md stubs by hand.

5. Make examples/example_step.py load a registered demo dataset with load_demo, call the library and assert on the result.

6. Verify it is indexed, then refresh the domain index: python -c "from omicsclaw.skills import load_skills;
   i = load_skills('skills'); print(len(i), i.skipped)" OMICSCLAW_WRITE_SKILL_INDEX=1 pytest
   tests/skills/test_domain_index_is_current.py pytest tests/sdk/notebook/test_skill_api_sections.py

Full usage notes and soft conventions live in templates/skill/README.md. -->

REPLACE_SKILL_NAME

When to use

<!-- One short paragraph (3-6 lines). Mirror the frontmatter description's load/skip split and explicitly call out the closest adjacent skill so the agent knows when to redirect. -->

The user has <input shape> and wants <output shape>. Pick this skill when <distinguishing condition>. For <adjacent capability> use <sibling-skill> instead.

Use from a step

library = load_skill("REPLACE_SKILL_NAME")
frame = read_input("data/<input>.csv")
write_output(library.run_method(frame, method="default"), "tables/replace_me.csv")

A complete step that runs on demo data: examples/example_step.py.

API

<!-- api:begin generated from _api.py; regenerate with run.py api <skill dir> --write -->

run_method(frame: pd.DataFrame, *, method: str='default') -> pd.DataFrame

Placeholder computation: return a copy of the table with a method column.

:param frame: The input table. :param method: The backend to run. Default "default", the only one so far; say here where each default comes from and when to change it. :returns: A new DataFrame: frame plus a method column. :raises ValueError: method is not one of METHODS.

<!-- api:end -->

Methods and parameters

<!-- Which method when. For every default: its value and where it comes from (the method's paper, the library's default, a lab convention). Which values the user should decide, and how to ask. -->

• method="default": the only backend so far.

Gotchas

<!-- The failure modes an agent hits, each with the fix. Anchor each to something grep-able: a function name, a quoted error message, a column or key. -->

• None yet — append as failure modes are reported.

Inputs and outputs

<!-- What each function reads (columns, obs/obsm/layers keys) and what it writes or returns. -->

• run_method reads a table and returns a new one with a method column.

CLI

replace_me.py runs the same functions outside a project and writes tables/replace_me.csv, report.md and result.json: python <skill directory>/replace_me.py --help. --demo synthesises its input.

Dependencies

Python packages this skill's script needs. They are not installed for you — check before a long run.

<package-a>, <package-b>

See also

• references/parameters.md — every CLI flag, per-method tunables
• references/methodology.md — the WHY behind the algorithm
• references/output_contract.md — tables/X.csv + result.json schema
• Adjacent skills: <sibling-1> (upstream), <sibling-2> (parallel), <sibling-3> (downstream)