implementation-validation-cycle
Skill17 starsChanged 28 days agoMITView on GitHub
Run the repository's feature-branch to orphan-test-branch validation lifecycle with a verified full SHA, remote GitHub Actions evidence, failure classification, PR reporting, and safe cleanup. Use when preparing or revalidating a feature PR, when a temporary remote validation workflow is needed, or when a validation-cycle failure must be diagnosed.
In this repository5 other files it gives its agents.
- AGENTS.mdAGENTS.md
- git-commit-helperSkill
- hermes-agent-env-ruleSkill
- release-flowSkill
- review-upstreamSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
3
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
- Very specific: line 8
Use this skill after the implementation has passed its required local checks and the user has approved the implementation commit.
- Distinctive: line 18
Run the preflight helper. It rejects abbreviated SHAs and a feature branch whose remote tip is not exactly the requested commit.
- Very specific: line 33
Use create without --apply first.
- Very specific: line 62
Only after the comment succeeds, inspect the exact test worktree then call cleanup --apply --comment-url <comment-url>.
- Distinctive: line 77
Add a profile only when its commands and assertions are deterministic.
≈ “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.
Files
Implementation validation cycle
Use this skill after the implementation has passed its required local checks and
the user has approved the implementation commit. It produces remote evidence;
it never changes the merge path from feat/<scope> to main.
Inputs and gates
Require all of these before mutation: implementation worktree, feature branch,
issue/PR number, validation profile, and the full implementation SHA from
git rev-parse HEAD. Push the implementation first.
Run the preflight helper. It rejects abbreviated SHAs and a feature branch whose remote tip is not exactly the requested commit.
scripts/validation-cycle.sh preflight \
--repo /path/to/repo --feature-branch feat/<scope> --sha <40-char-sha>For a chart change, local evidence normally includes make docs when
values.yaml changed, make lint, make template, make package, and an
isolated kind install plus helm test. Record an unavailable local dependency
honestly; do not claim the remote result was local evidence.
Remote loop
- Use
createwithout--applyfirst. Inspect the feature SHA, test branch, worktree path, profile, and generated workflow path. - With explicit approval, rerun
create --apply. It creates an orphantest/<scope>worktree containing only.github/workflows/validate-implementation.yaml, commits it, and pushes it. The workflow checks out the full pinned SHA. Capture the exact run ID and wait only for that run:
gh run watch <run-id> --repo <owner/repo> --exit-status- On failure, classify before editing:
- implementation: feature checkout, assertion, or chart workload fails. Return to the feature worktree, rerun local checks, commit a feature fix, and restart with its new SHA.
- harness: the orphan workflow, its pinned SHA, or its setup fails before testing the implementation. Change only the orphan branch and rerun.
- infrastructure: runner, registry, or transient platform failure. Preserve state and retry only when the failure is demonstrably transient; otherwise ask the maintainer.
On success, post one PR comment with the tested SHA, run URL, inputs, expected and actual results, and a short secret-safe log excerpt. Use the helper to print the Markdown skeleton:
scripts/validation-cycle.sh evidence \ --sha <40-char-sha> --run-url <actions-url> --input '<test input>' \ --expected '<expected result>' --actual '<actual result>'- Only after the comment succeeds, inspect the exact test worktree then call
cleanup --apply --comment-url <comment-url>. Never open a test-to-mainPR, merge the feature branch into the test branch, auto-merge a PR, or run a release.
Helpers
scripts/validation-cycle.sh: preflight, orphan-worktree creation, workflow rendering, evidence skeleton, and guarded cleanup.assets/validate-implementation.yaml.tmpl: the temporary chart-validation workflow. It checks out__IMPLEMENTATION_SHA__and is the only file copied to an orphan test branch.scripts/test-validation-cycle.sh: local fixture tests for the SHA mismatch, dry-run, workflow rendering, and evidence formatting paths.
Add a profile only when its commands and assertions are deterministic. Keep feature-specific assertions in the temporary workflow, not the implementation branch. See CONTRIBUTING.md's "Promoting a validation to permanent CI" for the criteria deciding whether a proven assertion should become permanent CI.
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
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.

