Content
96%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is an exceptionally lean, command-driven governance procedure: every step is executable, every write is gated by an explicit re-read or eligibility validation, and failure paths consistently stop rather than guess. Its only structural weakness is that everything lives in one ~170-line file where the activation-card protocol and comment template could be externalized as bundled references.
Suggestions
Move the activation-card template, its field rules, and the exit receipt into a bundled reference file (e.g. references/activation-card.md) and keep a one-line pointer in SKILL.md, since this protocol appears self-contained and likely shared across autopilot skills.
Extract the exact apm-scope:v1 comment format into a short reference file so SKILL.md reads as an overview of the workflow with the template loaded only at post time.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is telegraphic imperative prose with zero explanation of concepts Claude already knows (no explanation of gh, GitHub labels, or what a scope is) and no padding: "Do not implement the issue. Do not assign. Do not request reviewers." Only one or two safety-motivated duplicated constraints ("do not edit the comment" restated in the failure-handling sentence) could be trimmed, which fits the 5 anchor ('every token earns its place') better than the 4 anchor's 'minor instances of over-explanation'. | 5 / 5 |
Actionability | Quotes: "gh api user --jq '{login:.login,type:.type}'", "gh issue comment \"$ISSUE\" --repo microsoft/apm --body-file \"$BODY\"\"", "node scripts/governance/eligibility.cjs --repo microsoft/apm --issue \"$ISSUE\" --approval-url \"$COMMENT_URL\"\"", "gh issue edit \"$ISSUE\" --repo microsoft/apm --add-label status/accepted". All commands are copy-paste ready, the comment body is given as an exact fence template, variables are anchored ("`COMMENT_URL` is its `html_url`") and edge cases (missing brief, conflict, unclear Area, changed field) each have concrete handling — matching the 5 anchor. | 5 / 5 |
Workflow Clarity | Quotes: the sequence activation card → Draft → Chat approval → Identity → "Post, re-read, then label"; validation checkpoints include the four-bullet re-read checklist ("the body is the approved text", "user.login is the `gh api user` login", "updated_at equals created_at"), the eligibility gate "Require `state: record-present`", and stop conditions ("If that brief is missing, stop", "Any other token, bot, or login -> stop"). A feedback loop exists for changed drafts ("A change to any field... is a new draft. Show the new fence and wait again"), matching the 5 anchor; the destructive/batch cap does not apply because validation is pervasive. | 5 / 5 |
Progressive Disclosure | Sections are clearly headed and linearly navigable, and external dependencies are well signaled with provenance rules ("Read the single `apm-scope-roster` block in `GOVERNANCE.md` on the trusted default branch", eligibility tool "from a trusted default-branch checkout"). However, at ~170 lines in a single file it exceeds the under-50-line case where organized sections alone earn a 5, and the activation-card protocol (~40 lines of card template, rules, and exit receipt) and the exact comment-format spec are self-contained chunks that could be split into bundled reference files — minor organization gaps matching the 4 anchor rather than the 5 anchor's 'content appropriately split' with one-level-deep references. | 4 / 5 |
Total | 19 / 20 Passed |