agentleFS
qf-studioqf-studio/navigator/skills/nav-release/SKILL.md

nav-release

Skill355 starsChanged 13 months agoMITView on GitHub

Validate and release Navigator plugin to marketplace. Auto-invoke when user says "release plugin", "publish navigator", "create release", or "deploy new version".

  • Deletes or force-pushes
  • Commits and pushes

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    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:

    After v5.1.0 incident where nav-profile was referenced in plugin.json but never committed, causing install failures.

    line 14
  2. Distinctive:

    All referenced skills exist and are committed

    line 15
  3. Very specific:

    All skills in plugin.json exist

    line 43
  4. Distinctive:

    All skill files are committed (not untracked)

    line 44
  5. Distinctive:

    No uncommitted changes in skills/

    line 46
  6. Distinctive:

    If validation fails: STOP and fix issues before proceeding.

    line 48
  7. Very specific:

    If --verify-hooks fails: STOP. The hook command in plugin.json is broken in the CLAUDE_PLUGIN_ROOT-unset case.

    line 81
  8. Very specific:

    If the workflow fails, inspect with gh run view --log-failed and fix before retrying — do NOT fall back to gh release create locally.

    line 172

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Navigator Release Skill

Validate plugin integrity and release to marketplace with all safety checks.

Why This Exists

After v5.1.0 incident where nav-profile was referenced in plugin.json but never committed, causing install failures. This skill ensures:

  • All referenced skills exist and are committed
  • Version consistency across all files
  • Tag created AFTER all commits
  • Post-release verification

When to Invoke

Auto-invoke when:

  • User says "release plugin", "publish navigator"
  • User says "create release", "deploy new version"
  • User says "release vX.Y.Z"

DO NOT invoke if:

  • Just committing changes (no release)
  • Updating documentation only
  • Testing locally

Execution Steps

Step 1: Pre-Release Validation [CRITICAL]

Run validation script:

python3 functions/release_validator.py --check-all

This validates:

  1. All skills in plugin.json exist
  2. All skill files are committed (not untracked)
  3. Version consistency across files
  4. No uncommitted changes in skills/

If validation fails: STOP and fix issues before proceeding.

Step 1.5: Hook Smoke Test [STRONGLY RECOMMENDED for any release touching hooks or plugin.json]

Run hook smoke-test:

python3 functions/release_validator.py --verify-hooks

This executes every plugin manifest hook command via bash twice — once with $CLAUDE_PLUGIN_ROOT bound to the repo under release, once with it unset and HOME pointed at a tmp dir whose .claude/plugins/marketplaces/navigator-marketplace links to the repo, so the manifest fallback path is what resolves (TASK-99). NAVIGATOR_MOD_OWNS and PILOT_EXECUTOR are scrubbed from both runs; inside a Claude Code session the mod's ownership list would otherwise make the Python dispatcher exit silently and fake the v6.14.0 signature. It detects the v6.14.0 silent-fail signature: a payload-emitting hook (SessionStart, PreCompact, PostCompact) that exits 0 with no stdout and no stderr.

Other hook events (Stop, UserPromptSubmit, PreToolUse, PostToolUse) are silent by design — they're state-writers or blocking-only — so quiet exit 0 there is correct behavior and is not flagged.

Expected output on a healthy release:

Hook smoke-test: 20/20 passed, 0 failed

  ✓  SessionStart       [set  ] exit=0 out=10353B err=0B
  ✓  SessionStart       [unset] exit=0 out=10353B err=0B
  ✓  PreCompact         [set  ] exit=0 out=2B err=87B
  ✓  PreCompact         [unset] exit=0 out=2B err=87B
  ...

Regression output (what v6.14.0 would have shown):

  ❌ SessionStart       [unset] silent exit 0 — payload-emitting hook produced no output
     python3 "${CLAUDE_PLUGIN_ROOT:-...}/hooks/nav_session_start.py"

If --verify-hooks fails: STOP. The hook command in plugin.json is broken in the CLAUDE_PLUGIN_ROOT-unset case. Fix the command (typically a fallback-path expansion issue) and re-run before proceeding.

Also run the two static runtime gates (both are CI steps in release.yml since TASK-99):

python3 functions/release_validator.py --verify-dispatcher          # every manifest hook routes through nav_dispatch.py; every registry op has a committed, importable hooks/ops/<name>.py
python3 functions/release_validator.py --verify-conformance           # a harness-conformance results file exists for `claude --version` (CI passes the pinned version)

Background: this check was added in v6.15.2 after v6.14.0 shipped a shell guard (if [ -n "$CLAUDE_PLUGIN_DIR" ]; then ... fi) that silently no-opped every hook when the variable was unset. The bug masked itself for two releases because the navigator source repo had a project-local .claude/settings.json backstop. Other Nav-initialized projects (no backstop) got zero injection with zero error signal. See mem-036 and releases/RELEASE-NOTES-v6.15.1.md / v6.15.2.md.

Step 2: Display Validation Results

Show validation summary:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
NAVIGATOR RELEASE VALIDATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Skills Check:
  [x] nav-loop          ✓ exists, committed
  [x] nav-profile       ✓ exists, committed
  [x] nav-diagnose      ✓ exists, committed
  ...

Version Check:
  plugin.json:      5.1.0 ✓
  marketplace.json: 5.1.0 ✓
  CLAUDE.md:        5.1.0 ✓
  README.md:        5.1.0 ✓

Git Status:
  Uncommitted skills: 0 ✓
  Untracked skills:   0 ✓

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
VALIDATION: PASSED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Step 3: Confirm Version

Ask user to confirm version:

Ready to release Navigator vX.Y.Z

This will:
1. Commit any pending changes
2. Push to origin/main
3. Create and push git tag vX.Y.Z
4. CI workflow (.github/workflows/release.yml) publishes the GitHub release from the tag

Proceed? [Y/n]

Step 4: Execute Release

If confirmed, run release sequence:

# 1. Commit if needed
git add .
git commit -m "chore(release): prepare vX.Y.Z" || true

# 2. Push to origin
git push origin main

# 3. Create and push tag — this triggers .github/workflows/release.yml,
#    which creates the GitHub release from releases/RELEASE-NOTES-vX.Y.Z.md.
git tag -a vX.Y.Z -m "Navigator vX.Y.Z: [description]"
git push origin vX.Y.Z

Do NOT run gh release create locally. The workflow owns release publication. Local creation races the workflow and historically caused "release with the same tag name already exists" CI failures.

Step 5: Post-Release Verification

Wait for the Publish Release workflow and verify it succeeded:

# Watch the most recent run; exits non-zero if it fails.
gh run watch --exit-status "$(gh run list --workflow=release.yml --limit=1 --json databaseId --jq '.[0].databaseId')"

# Confirm the release is published
gh release view vX.Y.Z --json tagName,name,isDraft,assets

If the workflow fails, inspect with gh run view --log-failed and fix before retrying — do NOT fall back to gh release create locally.

Verify tag contains all skills:

python3 functions/release_validator.py --verify-tag vX.Y.Z

Clear local cache (for testing):

rm -rf ~/.claude/plugins/cache/navigator-marketplace/

Instruct user to test:

Release complete! To verify:

1. Run: /plugin install navigator
2. Check plugin list shows vX.Y.Z
3. Verify no errors in plugin details

If errors occur, see: .agent/sops/deployment/plugin-release.md

Step 6: Update Installed Plugin + Sync Docs Site

claude plugin update navigator && claude plugin list | grep -A1 navigator@   # then restart

The docs site is a separate repo (qf-studio/navigator-docs, local ~/Projects/startups/navigator-site, live at https://navigator.quantflow.studio on AWS); it does not update itself. Bump lib/version.ts, add/adjust content/skills/*.mdx and content/reference/nav-config-schema.mdx for anything new, bun run build, commit and push main, then push a prod-<VERSION>-$(date +%s) tag: "Build image" publishes the image to GHCR and "Deploy QuantFlow AWS" rolls it to ECS and invalidates CloudFront. Confirm the live version string at navigator.quantflow.studio. Full procedure: .agent/sops/development/release-workflow.md Step 7.

Predefined Functions

functions/release_validator.py

Validates plugin integrity before release:

# Check all skills exist and are committed
python3 functions/release_validator.py --check-all

# Verify specific version
python3 functions/release_validator.py --check-version 5.1.0

# Verify tag contents
python3 functions/release_validator.py --verify-tag v6.15.2

# Smoke-test plugin manifest hook commands (v6.15.2+)
python3 functions/release_validator.py --verify-hooks

# Python fallback runtime: dispatcher routing + registry ops exist, committed, import (TASK-99)
python3 functions/release_validator.py --verify-dispatcher

# Harness-conformance results exist for a Claude Code version (default: `claude --version`)
python3 functions/release_validator.py --verify-conformance 2.1.287

Error Handling

Missing skill detected:

❌ VALIDATION FAILED

Missing skills:
  - skills/nav-profile/ (referenced in plugin.json but not found)

Fix: Create the skill or remove from plugin.json

Uncommitted skills detected:

❌ VALIDATION FAILED

Uncommitted skills:
  - skills/nav-loop/ (modified)
  - skills/nav-profile/ (untracked)

Fix: git add skills/ && git commit -m "Add missing skills"

Version mismatch detected:

❌ VALIDATION FAILED

Version mismatch:
  plugin.json:      5.1.0
  marketplace.json: 5.0.0  ← MISMATCH
  CLAUDE.md:        5.1.0

Fix: Update marketplace.json to 5.1.0

Success Criteria

Release is successful when:

  • All skills validated (exist + committed)
  • Version consistent across all files
  • Hook smoke-test passes (--verify-hooks reports zero failures)
  • Git tag created after all commits
  • Publish Release workflow run completes successfully
  • GitHub release published with notes asset attached
  • Test installation succeeds (no errors)

Quick Reference

# Full release with validation
"Release Navigator v5.2.0"

# Just validate (no release)
"Validate plugin for release"

# Fix after failed release
"Fix release tag v5.1.0"

Related

  • SOP: .agent/sops/deployment/plugin-release.md
  • Config: .claude-plugin/plugin.json
  • Marketplace: .claude-plugin/marketplace.json

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.