agentleFS
Sign inStart free
tardigrdetardigrde/obsidian-agentic-chat/.agents/skills/ship-change/SKILL.md

Ship Change

Skill8 starsChanged 7 days agoMITView on GitHub

Use for any code change in this repo: bug, feature, refactor, or dependency upgrade. Routes the work, requires a witness test, and runs the lowest verify lane that can actually fail.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 6 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Witness first: Every change has the smallest check that fails before and passes after.

    line 12
  2. Very specific:

    Lowest failing lane: Run the cheapest tier that can actually fail for the risk you introduced.

    line 14
  3. Distinctive:

    UI claims need more than a screenshot: Use a committed e2e assertion, or a probe recording plus DOM/state metrics.

    line 18
  4. Distinctive:

    UI entry points need clips: If the change adds, removes, or changes a command, button, modal, settings row, mode, tab, or panel, plan …

    line 20
  5. Very specific:

    Read the matching file in .agents/skills/verify-obsidian-chat/features/.

    line 39
  6. Distinctive:

    Update the feature map: entry points, proof command, last verified date.

    line 53
  7. Distinctive:

    If the repro is not reproducible, tighten the conditions or instrument until it fires.

    line 60
  8. Distinctive:

    Prefer no test over a bad test.

    line 93

≈ “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

Ship change

Own the change end to end. Plan, implement, verify, and report evidence.

Non-negotiables

  1. Witness first. Every change has the smallest check that fails before and passes after. For a bug, run it before the fix and keep both outputs.
  2. Lowest failing lane. Run the cheapest tier that can actually fail for the risk you introduced. Do not run dogfood for a CSS change; do not ship a prompt change on unit tests alone.
  3. Counts, not exit codes. Zero tests or all tests skipped is a failure.
  4. UI claims need more than a screenshot. Use a committed e2e assertion, or a probe recording plus DOM/state metrics.
  5. UI entry points need clips. If the change adds, removes, or changes a command, button, modal, settings row, mode, tab, or panel, plan desktop 1280x1024 and mobile-width 390x844 clips before code. Each clip needs a "what to watch" list. A waiver must be explicit in the PR. The 390x844 clip is a narrow desktop window, not real Obsidian Mobile.
  6. No secret or vault data in a commit.

Route

requestroute
Bug reportReproduce on the matching surface first. Deterministic replay or e2e, not a guess. Then bisect, fix, and re-run the repro.
New or changed behaviorWrite the feature-map entry and name the entry points before code. Then the smallest implementation plus the witness.
RefactorBehavior preserving. Capture the baseline, change shape, prove the same checks still pass.
Dependency or Obsidian upgradeCapture the baseline, bump, prove the resolved version, then re-run behavior checks.
Idea with unknown behaviorDo not implement to spec. Run @probe-feature, decide from evidence, then come back.

Feature loop

  1. Read the matching file in .agents/skills/verify-obsidian-chat/features/.
  2. List every user entry point the change touches. Slash command, palette command, settings tab, composer control, drag and drop, agent tool, subagent. For each changed entry point, decide the desktop and mobile-width clips now, or record the waiver.
  3. Name the witness: the smallest assertion that would fail today.
  4. Implement the smallest change that satisfies it.
  5. Verify:
    • logic: npm run verify -- change <file-or-name>
    • types, lint, unit, bundle: npm run verify -- ci
    • merge gate: npm run verify -- fast
    • UI wiring: npm run verify -- e2e <spec> or npm run probe -- --scenario <name>
    • prompt, tool, or context: npm run verify -- eval plus compare --fail-on-regression
  6. Update the feature map: entry points, proof command, last verified date.
  7. Report evidence: command, exit code, counts, artifact path.

Bug loop

  1. Reproduce on the same surface a user would use. If it is UI, drive the real Obsidian app; if it is agent behavior, use test/helpers/agent-replay.ts.
  2. If the repro is not reproducible, tighten the conditions or instrument until it fires. Do not ship a "might fix" change.
  3. Stage the failing repro before the fix in git history.
  4. Fix the root cause. A guard is a fix only when the state it handles is legitimate.
  5. Re-run the repro. Unit tests show branch behavior, not bug absence.

Upgrade loop

  1. Baseline: npm run verify -- fast and record the result.
  2. Bump. Prove the resolved version with npm ls <package> and a lockfile diff.
  3. Re-run fast. If prompt, tool, context, or provider code changed, also run the eval A/B with --fail-on-regression.
  4. For Obsidian: also run npm run verify -- audit and the mobile lane.

Done means

  • Witness exists and was shown failing first.
  • The chosen lane is green with counts recorded.
  • Feature map updated if a user-facing entry point changed.
  • Changed entry points have desktop and mobile-width clips, or an explicit waiver.
  • Evidence is in the response, not just "it works".

Promotion

findingpromote to
Bug found livedeterministic replay test
New entry pointfeature map plus e2e assertion
Rule broken twicelint rule, script check, or CI gate
Layout regressionassertion in test/e2e/support/layout.ts
Flaky testfix the observation or fixture; never add a retry

Prefer no test over a bad test. A bad test mocks the subject, encodes the current implementation, or would be deleted immediately after proving the fix.

Read next

  • @verify-obsidian-chat for the real app and the feature map.
  • @probe-feature for exploratory work.
  • docs/development/verification.md for lanes and the fake-green list.

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.

Check your setup, free