[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6R6ACE612DH2JQCCB6V7QWV":3,"skill-same:01M6R6ACE612DH2JQCCB6V7QWV":69},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":43,"commitSha":44,"skillDirectory":45,"forksCount":13,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":48,"install":49,"resources":62},"01M6R6ACE612DH2JQCCB6V7QWV","risk-assessment",null,"Assess the delivery risk and ambiguity of an issue. Scores three anchored 1-5 risk dimensions — Code Blast Radius, Complexity, Business Blast Radius — into an overall risk level (Low\u002FMedium\u002FHigh) with safeguards, plus a separate Ambiguity score (how many open questions remain) with the open questions listed. Use this BEFORE work starts, whenever someone wants to know how risky, how big, how complex, or how well-specified an issue is, or asks to \"assess risk\", \"risk assessment\", \"score this issue\", \"how risky is this\", \"what's the blast radius\", \"how complex is this\", \"how ambiguous is this\", or \"how many open questions are there\". Accepts a Multica issue ID, a GitHub issue\u002FPR, or a free-text description as the argument.","diegoquinteiro\u002Fharnessfile","diegoquinteiro","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1878108?v=4","01M6R66DV4959S3MJ0BSC20JMG","01M6R66DYYMACQCANZQ4DFGEPR",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Frisk-assessment\u002FSKILL.md",[],[18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37],"assessment","risk","risk assessment","work","how complex","argument","free","as","delivery","three","dimensions","issue","complexity","text","how","open","radius","blast radius","code","pr","2026-11-10 10:12:30",false,"---\nname: risk-assessment\ndescription: >\n  Assess the delivery risk and ambiguity of an issue. Scores three anchored 1-5 risk dimensions —\n  Code Blast Radius, Complexity, Business Blast Radius — into an overall risk level (Low\u002FMedium\u002FHigh)\n  with safeguards, plus a separate Ambiguity score (how many open questions remain) with the open\n  questions listed. Use this BEFORE work starts, whenever someone wants to know how risky, how big,\n  how complex, or how well-specified an issue is, or asks to \"assess risk\", \"risk assessment\", \"score\n  this issue\", \"how risky is this\", \"what's the blast radius\", \"how complex is this\", \"how ambiguous\n  is this\", or \"how many open questions are there\". Accepts a Multica issue ID, a GitHub issue\u002FPR, or\n  a free-text description as the argument.\n---\n\n# Risk Assessment\n\nScore a single issue across three **risk** dimensions plus a separate **ambiguity** signal, and turn\nthem into an actionable verdict. This runs **early — before the change is planned or approved** — so\nboth the risk and the open questions are understood up front, not discovered after an incident.\n\n**Risk and ambiguity measure different things and are kept separate.** Risk (blast radius +\ncomplexity) is *how bad it is if the code is wrong* — a \"build it right\" signal. Ambiguity is *how\nlikely we are to build the wrong thing* — a \"build the right thing\" signal. A task can be high on one\nand low on the other, so ambiguity is **never folded into the risk number**; it is reported on its own.\n\n## Principles\n\n- **Reason per dimension, then score.** For each dimension, write the reasoning first and assign\n  the number last. A bare number with no justification is not acceptable.\n- **Anchor to evidence, not vibes.** Every score must cite the concrete files, modules, services,\n  or features that drive it. If you can't point to evidence, you haven't grounded the score.\n- **Use the anchored rubric.** Each level (1-5) has a named definition below. Match the issue to\n  the closest anchor; don't default to 3.\n- **List the open questions.** For ambiguity, don't just score it — enumerate the actual unresolved\n  questions a human would have to answer. The count and weight of those questions *is* the score.\n- **Judgment is yours; the math is deterministic.** You assign the four scores; `scripts\u002Fscore.py`\n  computes the overall risk level, the ambiguity band, and safeguards so the verdict is reproducible.\n- **Declare confidence.** For each dimension, note your confidence (high\u002Fmedium\u002Flow) and the single\n  biggest unknown that could move the score.\n\n## Step 1 — Resolve the input (auto-detect the source)\n\nThe argument is an issue identifier or a raw description. Detect which:\n\n- **Multica issue** — an issue ID\u002Fslug, or a `multica.ai` URL. Fetch it:\n  ```bash\n  multica issue get \u003Cid> --output json          # title, description, status, metadata\n  multica issue comment list \u003Cid> --output json # comments (may contain existing research)\n  ```\n  Set `source = multica`.\n- **GitHub issue\u002FPR** — a `#number`, an `owner\u002Frepo#number`, or a github.com URL. Fetch with\n  `gh issue view \u003Cn> --json title,body,comments` (or `gh pr view`). Set `source = github`.\n- **Free text** — anything else. Use the text as-is. Set `source = text`.\n\nIf the identifier is ambiguous or the fetch fails, tell the user what you tried and fall back to\ntreating the argument as free text.\n\n## Step 2 — Gather grounding context (reuse research if it already exists)\n\n**First check whether research already exists — if so, do NOT re-explore the codebase:**\n\n- A `RESEARCH.md` in the repo root (written by the `\u002Fresearch` skill).\n- A research \u002F Planner-handoff comment already on the Multica issue (from `comment list` in Step 1).\n- Substantial code context already gathered earlier in this conversation.\n\nIf grounded research exists, read it and score from it.\n\n**Otherwise, run a scoped, read-only risk exploration** — focused on risk signals only, not a full\nresearch pass. Use the `Explore` agent or grep\u002Fread to determine:\n\n- **Affected surface:** which files\u002Fmodules a solution would touch (`server\u002F`, `web\u002F`, both?).\n- **Fan-out \u002F coupling:** who imports or calls the affected code — how far a change ripples.\n- **Test coverage:** do the touched areas have tests, and at what tier (mocked\u002FVCR\u002Fintegration)?\n- **Cross-cutting touches:** DB migrations, GraphQL schema changes (which force frontend codegen),\n  auth\u002Faccount\u002Ftenant propagation, shared job infrastructure.\n\nKeep it tight. The goal is enough evidence to anchor the scores, not an exhaustive map.\n\n**Ambiguity is best assessed last — as the residue.** Score it *after* research, against whatever\nopen questions research could not resolve. Research exists precisely to drive out ambiguity, so the\nambiguity score reflects the questions that genuinely still need a human, not everything that was\nonce unclear.\n\n## Step 3 — Score the dimensions\n\nFor each dimension, match the issue to the closest anchor, write the reasoning with cited evidence,\nthen record the 1-5 score, your confidence, and the top unknown. The first three are **risk**\ndimensions (they feed the risk number); **Ambiguity** is scored and reported separately.\n\n### Code Blast Radius — *how much of the code a solution will affect*\n\n| Score | Anchor |\n|---|---|\n| **1 — Pinpoint** | A single file\u002Fcomponent; no exported\u002Fshared symbols; isolated. Copy, one Tailwind class, one self-contained function. No migration, no schema change. |\n| **2 — Local** | A handful of files within one feature\u002Fmodule; limited fan-out; stays inside one app (server *or* web). No DB migration, no GraphQL schema change. |\n| **3 — Module-wide** | A whole feature or one GraphQL type end-to-end (schema + resolver + frontend query + codegen). Moderate fan-out; may include a backward-compatible migration; crosses server↔web in a contained way. |\n| **4 — Cross-cutting** | Shared infrastructure many features import (base models, shared jobs, core GraphQL types), OR a non-trivial\u002Fdestructive migration, OR a change rippling across many call sites. High fan-out across both apps. |\n| **5 — Systemic** | Foundational plumbing used almost everywhere — core data-model relations, the job\u002F`INSTANCE` framework, account\u002Ftenant propagation, middleware\u002Fcontextvar layer, or a sweeping multi-dozen-file refactor. Hard to bound. |\n\n### Complexity — *how complex the solution is*\n\n| Score | Anchor |\n|---|---|\n| **1 — Trivial** | Obvious one-step change following a known pattern; config\u002Fstyling\u002Fcopy; no real logic. High confidence. |\n| **2 — Simple** | Straightforward CRUD\u002Fcomponent on an existing pattern; minimal logic; no new concepts. Low uncertainty. |\n| **3 — Moderate** | Several components wired together; some non-trivial logic, or a new GraphQL type + job, or a known 3rd-party\u002Fasync path. A few decisions, but the path is clear. |\n| **4 — Complex** | Significant new logic, async\u002Fconcurrency coordination, consistency\u002Frace concerns, new data modeling, or an unfamiliar integration. Several unknowns; needs a real plan and careful testing. |\n| **5 — Very complex** | Deep\u002Fnovel problem — distributed or concurrent correctness, large interdependent refactor, ML\u002Fmatching\u002Fembedding logic, performance-critical paths, or major unknowns. Expect iteration; hard to estimate. |\n\n> **Watch for deceptively simple framings.** A short ask like \"clean up the bad data\", \"figure out why X is slow\u002Fbroken\", or \"fix the flaky Y\" hides its work in an *unknown root cause or messy state* — the effort is unbounded until you've diagnosed it, and the fix is often larger than the request implies. Score these by the likely effort to resolve the unknown, not the brevity of the ask: investigation, data-cleanup, and \"why is this broken\" debugging tasks routinely land at **3–4** even when stated in one line. (The flip side also holds — a request that *sounds* sweeping but names one concrete, well-scoped change stays low.)\n\n### Business Blast Radius — *how critical the affected business features are*\n\n| Score | Anchor |\n|---|---|\n| **1 — Cosmetic** | Internal tooling, docs, styling, copy. A failure is invisible or trivial to users. No data, revenue, or trust impact. |\n| **2 — Peripheral** | A non-core, low-traffic feature. A failure annoys a few users; the core product keeps working. Easily reversible; no data-integrity concern. |\n| **3 — Important feature** | A feature users rely on regularly (Brainstorm\u002FIdeas\u002FSearch, data ingestion inputs). A failure degrades the experience or blocks a workflow, but no data loss or cross-account exposure. |\n| **4 — Core product value** | The **core analysis pipeline** (AnalysisConfig → AnalysisReport → Topic, comment\u002Ftopic matching, report generation) or the data its outputs depend on. A failure breaks the product's main value or corrupts analysis for many users. |\n| **5 — Critical \u002F trust & safety** | **Auth & tenant isolation**, account\u002Fdata separation, security, or billing\u002Fdata-loss\u002Ftotal-outage surfaces. A bug risks cross-account leakage or catastrophic, possibly regulatory, harm — not just a broken feature. |\n\n### Ambiguity — *how many open questions survive research* (separate signal, not part of risk)\n\nEnumerate the actual unresolved questions first, then pick the anchor that matches their count and weight.\n\n| Score | Anchor |\n|---|---|\n| **1 — Crystal clear** | Fully specified. Acceptance criteria obvious, intended behavior unambiguous. Zero open questions. |\n| **2 — Minor unknowns** | One or two small decisions the team can safely make itself (a default value, a label). No human input needed. |\n| **3 — Some open questions** | A few decisions that benefit from product input but aren't blocking; reasonable defaults exist. A human *should* weigh in. |\n| **4 — Significantly underspecified** | Multiple open questions about scope, behavior, or intent. Building without answers risks building the wrong thing; a human *must* weigh in before committing to a plan. |\n| **5 — Wide open \u002F exploratory** | The problem itself is fuzzy (\"make onboarding better\"). Needs discovery and human framing before any plan; proceeding blind would almost certainly miss the mark. |\n\n## Step 4 — Compute the verdict (deterministic)\n\nRun the aggregation script with the four scores and the detected `--source`:\n\n```bash\npython .agents\u002Fskills\u002Frisk-assessment\u002Fscripts\u002Fscore.py \\\n  --complexity \u003C1-5> --code-blast \u003C1-5> --business-blast \u003C1-5> --ambiguity \u003C1-5> \\\n  --source \u003Cmultica|github|text>\n```\n\nIt returns JSON with `overall_score`, `risk_level` (Low\u002FMedium\u002FHigh via\n`0.5·business + 0.3·code + 0.2·complexity`; Low ≤2.3 · Medium ≤3.6 · High >3.6), `safeguards`, and a\nseparate `ambiguity` block (`score` + `band`: 1-2 low · 3 medium · 4-5 high). Ambiguity is **not** in\nthe risk formula. Use these verbatim — do not recompute by hand.\n\n## Step 5 — Deliver\n\nRender the report in this format:\n\n```markdown\n## Risk Assessment — \u003Cissue title or summary>\n\n| Dimension | Score | Why |\n|---|---|---|\n| Code Blast Radius | N\u002F5 | \u003Cone line + cited evidence> |\n| Complexity | N\u002F5 | \u003Cone line + cited evidence> |\n| Business Blast Radius | N\u002F5 | \u003Cone line + cited evidence> |\n\n**Overall risk: \u003CLow\u002FMedium\u002FHigh>** (\u003Coverall_score> — weighted 0.5·business + 0.3·code + 0.2·complexity)\n\n**Ambiguity: N\u002F5 (\u003Clow|medium|high>)** — \u003Cone line on how open-ended the issue is>\n\n**Open questions** (the unresolved questions behind the ambiguity score)\n- \u003Ceach open question a human would need to answer; \"none\" if fully specified>\n\n**Watch-outs**\n- \u003Ceach safeguard from the script>\n\n**Confidence & unknowns**\n- \u003Cper-dimension confidence + the biggest unknown that could move a score>\n```\n\n**Then deliver the output based on the source:**\n\n- **`source = multica`** — write the assessment back to the issue:\n  ```bash\n  # Human-readable report as a comment (preserves the markdown verbatim):\n  printf '%s' \"\u003Creport markdown>\" | multica issue comment add \u003Cid> --content-stdin\n\n  # Machine-readable scores as metadata (queryable by the PM when it routes the issue):\n  multica issue metadata set \u003Cid> --key risk_code_blast      --value \u003CN>\n  multica issue metadata set \u003Cid> --key risk_complexity      --value \u003CN>\n  multica issue metadata set \u003Cid> --key risk_business_blast  --value \u003CN>\n  multica issue metadata set \u003Cid> --key risk_level           --value '\"\u003CLow|Medium|High>\"'\n  multica issue metadata set \u003Cid> --key risk_ambiguity       --value \u003CN>\n  ```\n  Confirm to the user that the comment and metadata were written, and share the issue link.\n\n- **`source = github` or `source = text`** — print the report to the terminal only. Do not write\n  anywhere external.\n","# Risk Assessment\n\nScore a single issue across three **risk** dimensions plus a separate **ambiguity** signal, and turn\nthem into an actionable verdict. This runs **early — before the change is planned or approved** — so\nboth the risk and the open questions are understood up front, not discovered after an incident.\n\n**Risk and ambiguity measure different things and are kept separate.** Risk (blast radius +\ncomplexity) is *how bad it is if the code is wrong* — a \"build it right\" signal. Ambiguity is *how\nlikely we are to build the wrong thing* — a \"build the right thing\" signal. A task can be high on one\nand low on the other, so ambiguity is **never folded into the risk number**; it is reported on its own.\n\n## Principles\n\n- **Reason per dimension, then score.** For each dimension, write the reasoning first and assign\n  the number last. A bare number with no justification is not acceptable.\n- **Anchor to evidence, not vibes.** Every score must cite the concrete files, modules, services,\n  or features that drive it. If you can't point to evidence, you haven't grounded the score.\n- **Use the anchored rubric.** Each level (1-5) has a named definition below. Match the issue to\n  the closest anchor; don't default to 3.\n- **List the open questions.** For ambiguity, don't just score it — enumerate the actual unresolved\n  questions a human would have to answer. The count and weight of those questions *is* the score.\n- **Judgment is yours; the math is deterministic.** You assign the four scores; `scripts\u002Fscore.py`\n  computes the overall risk level, the ambiguity band, and safeguards so the verdict is reproducible.\n- **Declare confidence.** For each dimension, note your confidence (high\u002Fmedium\u002Flow) and the single\n  biggest unknown that could move the score.\n\n## Step 1 — Resolve the input (auto-detect the source)\n\nThe argument is an issue identifier or a raw description. Detect which:\n\n- **Multica issue** — an issue ID\u002Fslug, or a `multica.ai` URL. Fetch it:\n  ```bash\n  multica issue get \u003Cid> --output json          # title, description, status, metadata\n  multica issue comment list \u003Cid> --output json # comments (may contain existing research)\n  ```\n  Set `source = multica`.\n- **GitHub issue\u002FPR** — a `#number`, an `owner\u002Frepo#number`, or a github.com URL. Fetch with\n  `gh issue view \u003Cn> --json title,body,comments` (or `gh pr view`). Set `source = github`.\n- **Free text** — anything else. Use the text as-is. Set `source = text`.\n\nIf the identifier is ambiguous or the fetch fails, tell the user what you tried and fall back to\ntreating the argument as free text.\n\n## Step 2 — Gather grounding context (reuse research if it already exists)\n\n**First check whether research already exists — if so, do NOT re-explore the codebase:**\n\n- A `RESEARCH.md` in the repo root (written by the `\u002Fresearch` skill).\n- A research \u002F Planner-handoff comment already on the Multica issue (from `comment list` in Step 1).\n- Substantial code context already gathered earlier in this conversation.\n\nIf grounded research exists, read it and score from it.\n\n**Otherwise, run a scoped, read-only risk exploration** — focused on risk signals only, not a full\nresearch pass. Use the `Explore` agent or grep\u002Fread to determine:\n\n- **Affected surface:** which files\u002Fmodules a solution would touch (`server\u002F`, `web\u002F`, both?).\n- **Fan-out \u002F coupling:** who imports or calls the affected code — how far a change ripples.\n- **Test coverage:** do the touched areas have tests, and at what tier (mocked\u002FVCR\u002Fintegration)?\n- **Cross-cutting touches:** DB migrations, GraphQL schema changes (which force frontend codegen),\n  auth\u002Faccount\u002Ftenant propagation, shared job infrastructure.\n\nKeep it tight. The goal is enough evidence to anchor the scores, not an exhaustive map.\n\n**Ambiguity is best assessed last — as the residue.** Score it *after* research, against whatever\nopen questions research could not resolve. Research exists precisely to drive out ambiguity, so the\nambiguity score reflects the questions that genuinely still need a human, not everything that was\nonce unclear.\n\n## Step 3 — Score the dimensions\n\nFor each dimension, match the issue to the closest anchor, write the reasoning with cited evidence,\nthen record the 1-5 score, your confidence, and the top unknown. The first three are **risk**\ndimensions (they feed the risk number); **Ambiguity** is scored and reported separately.\n\n### Code Blast Radius — *how much of the code a solution will affect*\n\n| Score | Anchor |\n|---|---|\n| **1 — Pinpoint** | A single file\u002Fcomponent; no exported\u002Fshared symbols; isolated. Copy, one Tailwind class, one self-contained function. No migration, no schema change. |\n| **2 — Local** | A handful of files within one feature\u002Fmodule; limited fan-out; stays inside one app (server *or* web). No DB migration, no GraphQL schema change. |\n| **3 — Module-wide** | A whole feature or one GraphQL type end-to-end (schema + resolver + frontend query + codegen). Moderate fan-out; may include a backward-compatible migration; crosses server↔web in a contained way. |\n| **4 — Cross-cutting** | Shared infrastructure many features import (base models, shared jobs, core GraphQL types), OR a non-trivial\u002Fdestructive migration, OR a change rippling across many call sites. High fan-out across both apps. |\n| **5 — Systemic** | Foundational plumbing used almost everywhere — core data-model relations, the job\u002F`INSTANCE` framework, account\u002Ftenant propagation, middleware\u002Fcontextvar layer, or a sweeping multi-dozen-file refactor. Hard to bound. |\n\n### Complexity — *how complex the solution is*\n\n| Score | Anchor |\n|---|---|\n| **1 — Trivial** | Obvious one-step change following a known pattern; config\u002Fstyling\u002Fcopy; no real logic. High confidence. |\n| **2 — Simple** | Straightforward CRUD\u002Fcomponent on an existing pattern; minimal logic; no new concepts. Low uncertainty. |\n| **3 — Moderate** | Several components wired together; some non-trivial logic, or a new GraphQL type + job, or a known 3rd-party\u002Fasync path. A few decisions, but the path is clear. |\n| **4 — Complex** | Significant new logic, async\u002Fconcurrency coordination, consistency\u002Frace concerns, new data modeling, or an unfamiliar integration. Several unknowns; needs a real plan and careful testing. |\n| **5 — Very complex** | Deep\u002Fnovel problem — distributed or concurrent correctness, large interdependent refactor, ML\u002Fmatching\u002Fembedding logic, performance-critical paths, or major unknowns. Expect iteration; hard to estimate. |\n\n> **Watch for deceptively simple framings.** A short ask like \"clean up the bad data\", \"figure out why X is slow\u002Fbroken\", or \"fix the flaky Y\" hides its work in an *unknown root cause or messy state* — the effort is unbounded until you've diagnosed it, and the fix is often larger than the request implies. Score these by the likely effort to resolve the unknown, not the brevity of the ask: investigation, data-cleanup, and \"why is this broken\" debugging tasks routinely land at **3–4** even when stated in one line. (The flip side also holds — a request that *sounds* sweeping but names one concrete, well-scoped change stays low.)\n\n### Business Blast Radius — *how critical the affected business features are*\n\n| Score | Anchor |\n|---|---|\n| **1 — Cosmetic** | Internal tooling, docs, styling, copy. A failure is invisible or trivial to users. No data, revenue, or trust impact. |\n| **2 — Peripheral** | A non-core, low-traffic feature. A failure annoys a few users; the core product keeps working. Easily reversible; no data-integrity concern. |\n| **3 — Important feature** | A feature users rely on regularly (Brainstorm\u002FIdeas\u002FSearch, data ingestion inputs). A failure degrades the experience or blocks a workflow, but no data loss or cross-account exposure. |\n| **4 — Core product value** | The **core analysis pipeline** (AnalysisConfig → AnalysisReport → Topic, comment\u002Ftopic matching, report generation) or the data its outputs depend on. A failure breaks the product's main value or corrupts analysis for many users. |\n| **5 — Critical \u002F trust & safety** | **Auth & tenant isolation**, account\u002Fdata separation, security, or billing\u002Fdata-loss\u002Ftotal-outage surfaces. A bug risks cross-account leakage or catastrophic, possibly regulatory, harm — not just a broken feature. |\n\n### Ambiguity — *how many open questions survive research* (separate signal, not part of risk)\n\nEnumerate the actual unresolved questions first, then pick the anchor that matches their count and weight.\n\n| Score | Anchor |\n|---|---|\n| **1 — Crystal clear** | Fully specified. Acceptance criteria obvious, intended behavior unambiguous. Zero open questions. |\n| **2 — Minor unknowns** | One or two small decisions the team can safely make itself (a default value, a label). No human input needed. |\n| **3 — Some open questions** | A few decisions that benefit from product input but aren't blocking; reasonable defaults exist. A human *should* weigh in. |\n| **4 — Significantly underspecified** | Multiple open questions about scope, behavior, or intent. Building without answers risks building the wrong thing; a human *must* weigh in before committing to a plan. |\n| **5 — Wide open \u002F exploratory** | The problem itself is fuzzy (\"make onboarding better\"). Needs discovery and human framing before any plan; proceeding blind would almost certainly miss the mark. |\n\n## Step 4 — Compute the verdict (deterministic)\n\nRun the aggregation script with the four scores and the detected `--source`:\n\n```bash\npython .agents\u002Fskills\u002Frisk-assessment\u002Fscripts\u002Fscore.py \\\n  --complexity \u003C1-5> --code-blast \u003C1-5> --business-blast \u003C1-5> --ambiguity \u003C1-5> \\\n  --source \u003Cmultica|github|text>\n```\n\nIt returns JSON with `overall_score`, `risk_level` (Low\u002FMedium\u002FHigh via\n`0.5·business + 0.3·code + 0.2·complexity`; Low ≤2.3 · Medium ≤3.6 · High >3.6), `safeguards`, and a\nseparate `ambiguity` block (`score` + `band`: 1-2 low · 3 medium · 4-5 high). Ambiguity is **not** in\nthe risk formula. Use these verbatim — do not recompute by hand.\n\n## Step 5 — Deliver\n\nRender the report in this format:\n\n```markdown\n## Risk Assessment — \u003Cissue title or summary>\n\n| Dimension | Score | Why |\n|---|---|---|\n| Code Blast Radius | N\u002F5 | \u003Cone line + cited evidence> |\n| Complexity | N\u002F5 | \u003Cone line + cited evidence> |\n| Business Blast Radius | N\u002F5 | \u003Cone line + cited evidence> |\n\n**Overall risk: \u003CLow\u002FMedium\u002FHigh>** (\u003Coverall_score> — weighted 0.5·business + 0.3·code + 0.2·complexity)\n\n**Ambiguity: N\u002F5 (\u003Clow|medium|high>)** — \u003Cone line on how open-ended the issue is>\n\n**Open questions** (the unresolved questions behind the ambiguity score)\n- \u003Ceach open question a human would need to answer; \"none\" if fully specified>\n\n**Watch-outs**\n- \u003Ceach safeguard from the script>\n\n**Confidence & unknowns**\n- \u003Cper-dimension confidence + the biggest unknown that could move a score>\n```\n\n**Then deliver the output based on the source:**\n\n- **`source = multica`** — write the assessment back to the issue:\n  ```bash\n  # Human-readable report as a comment (preserves the markdown verbatim):\n  printf '%s' \"\u003Creport markdown>\" | multica issue comment add \u003Cid> --content-stdin\n\n  # Machine-readable scores as metadata (queryable by the PM when it routes the issue):\n  multica issue metadata set \u003Cid> --key risk_code_blast      --value \u003CN>\n  multica issue metadata set \u003Cid> --key risk_complexity      --value \u003CN>\n  multica issue metadata set \u003Cid> --key risk_business_blast  --value \u003CN>\n  multica issue metadata set \u003Cid> --key risk_level           --value '\"\u003CLow|Medium|High>\"'\n  multica issue metadata set \u003Cid> --key risk_ambiguity       --value \u003CN>\n  ```\n  Confirm to the user that the comment and metadata were written, and share the issue link.\n\n- **`source = github` or `source = text`** — print the report to the terminal only. Do not write\n  anywhere external.\n",{"name":5,"description":7},"main","c24d356df2a247ed3d08650881e768f30fdb293d","ui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Frisk-assessment","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro","2026-09-28 11:26:35",{"source":8,"skillName":5,"commands":50,"prompt":60,"zipAvailable":61},[51,54,57],{"tool":52,"command":53},"npx","npx skills add diegoquinteiro\u002Fharnessfile --skill risk-assessment",{"tool":55,"command":56},"bunx","bunx skills add diegoquinteiro\u002Fharnessfile --skill risk-assessment",{"tool":58,"command":59},"pnpm","pnpm dlx skills add diegoquinteiro\u002Fharnessfile --skill risk-assessment","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Frisk-assessment\u002FSKILL.md",true,[63],{"path":64,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":67,"sourceHtmlUrl":68},"scripts\u002Fscore.py","SCRIPT","脚本",4057,"https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Frisk-assessment\u002Fscripts\u002Fscore.py",[70,94,124,146,169,182],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":38},"01M6R6AREK1HJ2354SCCVK0MNS","wireframe","Produce low-fidelity wireframes or storyboards to explore a flow, layout, or idea across multiple options before committing to hi-fi design. Use when the user wants to 'explore options', 'sketch something out', 'see a few directions', or when the problem is open enough that hi-fi work would be wasted.","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Fwireframe\u002FSKILL.md",[],[72,77,78,79,33,80,21,81,82,83,84,85,86,87,88,89,90,91,92,93],"design","produce","flow","explore","user","use","sketch","idea","options","across","layout","wireframes","hi","problem","fi","directions","when",{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":108,"publishedAt":38},"01M6R6AJZ4MMXR6K89WZXAAK4Q","video-test","Record a video test of any feature or user flow via Playwright. Use when the user says \"record video test\", \"video test\", \"record a test\", \"show me the feature\", \"record the UI\", \"capture video\", or wants a visual demo of a feature. Accepts a description of what to test as an argument (e.g., `\u002Fvideo-test the brainstorm creation flow`).","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Fvideo-test\u002FSKILL.md",[100,105],{"categoryNo":101,"categoryName":102,"taxonomyType":103,"taxonomyTypeDesc":104},"01JBSEEDCAT000000000000008","测试与质量","SCENARIO","使用场景",{"categoryNo":106,"categoryName":107,"taxonomyType":103,"taxonomyTypeDesc":104},"01JBSEEDCAT000000000000016","图像与多媒体",[109,110,111,112,113,114,79,115,116,82,117,118,119,120,121,122,25,123,23,93],"test","Video","video test","playwright","record video","demo","show me","brainstorm","show","record","me","visual","creation","feature","via",{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":128,"categories":129,"keywords":133,"publishedAt":38},"01M6R6A0H1PXVJ5YGAR0GQPFTP","research","Deep parallel codebase research before planning or implementation. Spawns 4 specialized agents that explore behavior, architecture, coding standards, and test coverage simultaneously, then synthesizes findings into a structured RESEARCH.md document. Use this skill when the user says \"research\", \"investigate\", \"explore the codebase\", \"understand the code\", \"deep dive\", or wants thorough analysis of an area before making changes. Also use it as the first step of any complex feature — research first, then plan, then implement.","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Fresearch\u002FSKILL.md",[130],{"categoryNo":131,"categoryName":132,"taxonomyType":103,"taxonomyTypeDesc":104},"01JBSEEDCAT000000000000021","资料检索与研究",[126,134,109,135,36,136,137,138,80,139,140,141,82,142,143,144,122,25,145,93],"implement","architecture","planning","analysis","investigate","parallel","specialized","coverage","standards","document","deep","deep dive",{"skillNo":147,"name":148,"nameZh":6,"description":149,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":150,"licenseFileUrl":6,"sourceHtmlUrl":151,"categories":152,"keywords":153,"publishedAt":38},"01M6R69XDJ6KYJE6JNEKBP4FAY","ponytail","Lazy senior dev mode for any coding task (write, refactor, fix, review): YAGNI, stdlib first, no unrequested abstractions. Not for non-coding requests.","MIT","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Fponytail\u002FSKILL.md",[],[148,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168],"review","github","dev","https","refactor","first","coding","write","fix","yagni","task","requests","mode","for","lazy",{"skillNo":170,"name":171,"nameZh":6,"description":172,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":150,"licenseFileUrl":6,"sourceHtmlUrl":173,"categories":174,"keywords":175,"publishedAt":38},"01M6R69TX91W2HT26WA7VYHYGP","ponytail-review","Review a diff for over-engineering. Finds what to delete: reinvented stdlib, needless deps, speculative abstractions. One line per finding.","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Fponytail-review\u002FSKILL.md",[],[154,176,148,155,177,157,178,179,180,181,167],"ponytail review","delete","engineering","deps","diff","line",{"skillNo":183,"name":184,"nameZh":6,"description":185,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":186,"categories":187,"keywords":191,"publishedAt":38},"01M6R69PMB2G6QDK39N9VVV6Z3","polish-pass","Run a comprehensive end-of-design quality gate before a design is shown to stakeholders or shipped, covering hierarchy, AI-slop, interaction states, and accessibility. Use as the final review pass before delivery, or when the user asks to 'polish', 'finalize', or 'make it shippable'.","https:\u002F\u002Fgithub.com\u002Fdiegoquinteiro\u002Fharnessfile\u002Fblob\u002Fmain\u002Fui\u002Fsrc\u002Fdemo\u002Faltavox\u002Fskills\u002Fpolish-pass\u002FSKILL.md",[188],{"categoryNo":189,"categoryName":190,"taxonomyType":103,"taxonomyTypeDesc":104},"01JBSEEDCAT000000000000015","视觉设计",[192,193,77,154,194,195,196,197,82,198,199,26,200,25,201,202,203,204,205,93],"polish","polish pass","accessibility","quality gate","run","quality","gate","design is","end","interaction","it","states","slop","hierarchy"]