agentleFS
rioX432rioX432/agent-witness/.claude/skills/dev-all/SKILL.md

dev-all

Skill1 starsChanged 5 months agoApache-2.0View on GitHub

Process issues sequentially: /dev per issue in isolated sub-agent → CI wait → merge → next

  • Commits and pushes

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. Aurealibe/claude-config★ 68same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository12 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    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
  1. Very specific:

    Read the project's CLAUDE.md and look for ## Core Values section

    line 49
  2. Distinctive:

    If user chooses to proceed without, log a warning in the final report

    line 53
  3. Very specific:

    Check issue bodies for: blocked by #N, depends on #N, after #N

    line 86
  4. Distinctive:

    Ask user to confirm before proceeding.

    line 105
  5. Distinctive:

    Create a master task tracker:

    line 111
  6. Distinctive:

    After the sub-agent completes, validate the result before proceeding to merge:

    line 141
  7. Very specific:

    Read workspace/{issue}/review.json to get the structured review output

    line 143
  8. Distinctive:

    Ask: skip this issue and continue, or stop?

    line 168
  9. Very specific:

    When the user invokes /dev-all with /goal, the entire batch runs autonomously:

    line 194
  10. Distinctive:

    Stop only on 3 consecutive failures

    line 203

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “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

/dev-all — Sequential Issue Processing

Process multiple GitHub Issues sequentially. Each issue runs /dev in an isolated sub-agent, then waits for CI and merges before proceeding to the next.

Arguments: $ARGUMENTS

Why Per-Issue (not Single Branch)?

Each issue gets its own branch, PR, and merge cycle:

  • Clean git history: each PR is atomic and reviewable
  • CI validates each change independently
  • Merge conflicts are impossible: each issue starts from latest main
  • Rollback is easy: revert a single PR, not a batch

Step 0: Core Value Check (GATE)

  1. Read the project's CLAUDE.md and look for ## Core Values section
  2. If missing: Warn the user that Core Values are undefined. Ask if they want to:
    • Define Core Values now (recommended)
    • Proceed without the filter (not recommended — risk of feature bloat)
  3. If user chooses to proceed without, log a warning in the final report

Step 1: Resolve Target Issues

If $ARGUMENTS is provided: Extract issue numbers. If empty: Fetch all open issues:

gh issue list --state open --json number,title,labels,body --limit 100

1a. Filter Issues

  • Skip issues labeled won't — these are explicitly decided not to implement
  • Skip issues listed in CLAUDE.md ## Won't Do — cross-reference issue titles

Step 2: Parallel Investigation (Read-Only)

Launch parallel Explore agents (one per issue) to quickly understand scope:

Each agent:

  1. gh issue view {NUMBER} --json title,body,labels,comments
  2. Grep/Glob to find related code
  3. Return: summary, affected files, estimated scope, dependencies

Step 3: Dependency Analysis & Order

3a. Detect Dependencies

Check issue bodies for: blocked by #N, depends on #N, after #N

3b. Execution Order

Topological sort:

  1. Independent issues first (ascending by number)
  2. Dependent issues after their dependencies
  3. Circular dependencies → skip, report

── AskUserQuestion: Execution Plan ──

Present:

  1. Ordered list of issues
  2. Dependencies detected
  3. Skipped issues (with reasons — including won't label and Won't Do matches)
  4. Estimated scope per issue
  5. Core Value alignment per issue (if Core Values are defined)

Ask user to confirm before proceeding.


Step 4: Sequential Issue Loop

Create a master task tracker:

TaskCreate for each issue: "#{number}: {title}"

For each issue (in order):

4a. Pull latest main

git checkout main && git pull origin main

4b. Run /dev in isolated sub-agent (autonomous via /goal)

Agent(
  prompt: "/goal 'Issue #{issue_number} is resolved: tests pass, review has no Critical findings, and PR is created' /dev #{issue_number}",
  model: "opus",
  isolation: "worktree"
)

The sub-agent:

  • Gets a fresh context (no pollution from previous issues)
  • Works in an isolated git worktree (no file conflicts)
  • Runs the full /dev workflow autonomously via /goal
  • Skips AskUserQuestion confirmations (proceeds with best judgment)
  • Returns: structured result with PR URL, review status, and counts

4b-result. Review Validation

After the sub-agent completes, validate the result before proceeding to merge:

  1. Read workspace/{issue}/review.json to get the structured review output
  2. Parse the sub-agent's return value for review status

Decision logic:

Review StatusAction
critical (critical_count > 0)Skip this issue. Report to user: "#{issue} has {N} critical findings — skipping." Mark task as failed. Proceed to next issue.
warnings (warning_count > 0)Report to user. AskUserQuestion: "#{issue} PR has {N} unresolved warnings. Merge anyway?" If yes → proceed. If no → skip.
cleanProceed to auto-merge.
Sub-agent failed (status: "failed")Skip this issue. Report failure reason. Proceed to next issue.

4c. Enable auto-merge

gh pr merge {PR_URL} --auto --merge --delete-branch

4d. Wait for merge

Poll until merged (check every 30 seconds, timeout 15 minutes):

STATE=$(gh pr view {PR_URL} --json state -q '.state')

If CI fails:

  1. Report the failure to user
  2. Ask: skip this issue and continue, or stop?

4e. Mark task completed and proceed


Step 5: Final Report

## Batch Development Summary

| # | Issue | PR | Status |
|---|-------|----|--------|
| 1 | #{42} Title | PR_URL | Merged |
| 2 | #{43} Title | PR_URL | Merged |
| 3 | #{44} Title | — | Skipped (CI failed) |

Completed: N / M issues

Mark all tasks completed.


Autonomous Mode (/goal)

When the user invokes /dev-all with /goal, the entire batch runs autonomously:

/goal "All issues in $ARGUMENTS are resolved: each has a merged PR or a documented skip reason"

In autonomous mode:

  • Skip AskUserQuestion confirmations — proceed with best judgment
  • On CI failure: skip the issue and continue (don't stop)
  • Stop only on 3 consecutive failures

Error Handling

SituationAction
Issue not foundSkip, warn in report
Circular dependencySkip affected issues, report
Sub-agent /dev failsAsk user: skip or stop
CI failsAsk user: skip or stop
Merge conflictAsk user: skip or stop
3 consecutive failuresStop, report to user
Auto-merge timeout (15min)Report, ask user

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.