agentleFS
zajcazajca/pohunek/.agents/skills/milestone-review/SKILL.md

milestone-review

Skill9 starsChanged 3 days agoMITView on GitHub

Read-only review of a branch/worktree against its GitHub issue's definition-of-done, reporting each item with path:line evidence, then delegating any discrepancy fixes to a subagent and re-running the gates. Use when the user asks for a detailed review of a branch or to verify a milestone matches its spec.

19 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use soak-test, team-live-ops and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • lilangMaxSkill

    soak-test

    Generate a soak test protocol for extended play sessions. Defines what to observe, measure, and log during long play sessions to surface slow leaks, fatigue effects, and edge cases that only appear after sustained play. Primarily used in Polish and Release phases.

    Also in 9 projects that use this

  • lilangMaxSkill

    team-live-ops

    Orchestrate the live-ops team for post-launch content planning: coordinates live-ops-designer, economy-designer, analytics-engineer, community-manager, writer, and narrative-director to design and plan a season, event, or live content update.

    Also in 9 projects that use this

  • +10 more from lilangMax

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Review read-only, item by item: Check out or cd into the branch/worktree and verify each DoD item is actually implemented.

    line 37
  2. Very specific:

    Re-run the gates after fixes: Run the checks the fixes affect while iterating, then the gates skill on the branch's final revision.

    line 62
  3. Distinctive:

    Verify effects — do not report a discrepancy as fixed without re-reading the code and re-running the affected gate.

    line 81
  4. Very specific:

    Do not merge here; merging and advancing is the merge-advance skill's job.

    line 83
  5. Very specific:

    Follow the github-workflow skill's safe-persistence rules for all reads and writes to the issue/project.

    line 84

“Common” is measured: kept by at least 6 in 10 of the 18 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

milestone-review — verify a branch against its issue

Reviews an implemented milestone branch/worktree against the definition-of-done recorded on its GitHub issue, then drives fixes for any gap. This is the review half of the milestone loop: it does not implement scope itself, it verifies scope and delegates corrections. Findings and evidence are persisted on the issue via the github-workflow skill.

Inputs

  • The GitHub issue specifying the milestone (explicit URL/number from the user or task; otherwise infer per the github-workflow skill's resolution rules — unique unambiguous match only, otherwise ask).
  • The branch or worktree to review (e.g. zajca/milestone-4-attach-stream, or an absolute path like <primary-parent>/pohunek-worktrees/milestone-3-pty-sessions, where <primary-parent> is the directory holding the primary checkout). For a stack, the ordered slice branches/PRs and the slice plan recorded on the issue.

Steps

  1. Load the DoD. Fetch the live issue body and comments (comments may refine scope or record decisions) and enumerate every definition-of-done item with its stable ID. These are the checklist you review against — nothing more, nothing less. For a stack, also load the issue's slice plan (slice → DoD items).
  2. Review read-only, item by item. Check out or cd into the branch/worktree and verify each DoD item is actually implemented. For every item record a verdict (met / partial / missing) with concrete path:line evidence. Read the applicable .agents/rust-guidelines/ files so review comments match the repo's conventions (typed errors, no silent defaults, documented magic values, M-CANONICAL-DOCS, and the testing policy (.agents/instructions/verification.md#testing-policy) — a missing regression scenario for a bug or lifecycle fix is a gap, and a new unit test (an isolated component or per-function scenario, whatever file it lives in) is itself a finding; a missing unit test is not). For a deeper pass this maps well to parallel specialist reviewers (security-reviewer for the sh -c attach surface and secret handling, silent-failure-hunter for swallowed daemon errors, performance-reviewer), synthesized before you report.
  3. Record findings on the issue. Post the per-item verdict table with path:line evidence and the confirmed gaps to the issue as a comment (via the github-workflow skill). Keep the project status at In Progress while gaps are open; blockers and review state live in comments.
  4. Delegate the fixes. For each confirmed discrepancy, hand the fix to a subagent (this is the standing default). Brief each with concrete path:line context and the exact DoD item it must satisfy. Do not silently fix and re-review in one blur — keep the review findings and the fix work traceable on the issue. Verified out-of-scope findings discovered during the review go to their own follow-up issues (dedup via the github-workflow skill); an unmet original DoD item is never moved to a follow-up to claim the milestone done.
  5. Re-run the gates after fixes. Run the checks the fixes affect while iterating, then the gates skill on the branch's final revision. Iterate until green. For a stack, review each slice against its own diff (git diff <lower-slice-branch>...<slice-branch>) and the DoD items the slice plan assigns it, and flag a slice that mixes concerns, exceeds the pullRequests size cue in .github/agent-workflow.json without reason, or depends on a later slice (stub, dead code, placeholder, or red gates on its own). Fixes go to the slice that owns the gap, then the branches above it are restacked.
  6. Final verdict. Walk each DoD item with its final status and path:line evidence, and state the gate results. Say plainly whether the milestone matches its issue's specification, and record that final verdict as an issue comment too.

Constraints

  • The review pass itself is read-only; changes happen only through delegated fix work, then re-verification.
  • Verify effects — do not report a discrepancy as fixed without re-reading the code and re-running the affected gate.
  • Do not merge here; merging and advancing is the merge-advance skill's job.
  • Follow the github-workflow skill's safe-persistence rules for all reads and writes to the issue/project; do not close the issue or set Done here.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.