agentleFS
sgaabdu4sgaabdu4/building-flutter-apps/.agents/skills/he-build/SKILL.md

he-build

Skill25 starsChanged 3 months agoMITView on GitHub

Implement and verify a ready, authorized plan through focused fixes and integrated checks, ending with a ready-for-ship handoff. Use for implementation, bug fixes or resuming build work; skip planning-only, review-only and shipping-only requests.

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use isolation when needed.

    line 41

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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

Hard Eng Build

  • Baseline (checks before your edits) fails → repair only that, as its own change + PR; no feature edits until it merges (repair route).
  • Input = Ready + authorized plan from HE Plan; check its evidence (passing baseline, UX, blockers), not Status alone. Other missing, pending or contradictory readiness → that owner before production edits. Material scope change → same owner; preserve accepted decisions + completed work. Participation and project context → HE workflow.
  • Output = locally implemented + verified behavior, Complete plan (none for a change check rates Size: small) and explicit Ready for ship handoff. Build adds no authority to commit, push, publish, merge or deploy; delivery is a separate stage.

Implement + verify

flowchart TD
  B[Next complete behavior] --> T[Implement + focused checks]
  T --> R[Review actual diff + affected proof]
  R --> V{Pass?}
  V -->|No| F[Diagnose + fix owner]
  F --> T
  V -->|Yes| N{Remaining build work?}
  N -->|Yes| B
  N -->|No| I[Integrated proof + Complete gate]
  I -->|Fail| F
  I -->|Pass| H[Ready for ship]
  click T "../he/references/testing.md"
  click R "../code-review/SKILL.md"
  click F "../research/references/troubleshooting.md"
  click I "../he/references/gates.md#plan-checks"
  • Behavior = required connected callers, persistence, API and interface work; a skeleton or file checklist is not an accepted outcome. Keep changes at existing owners; apply relevant stack/design/security guidance only for the changed boundary.
  • Proof = test quality + actual-diff review; outside service → fake at its boundary from real responses, tests never call the real one; required runtime journeys, visual evidence and applicable accessibility states → E2E. PR evidence selection → HE Ship. Reuse these owners for regression, defect reopening and review findings; no duplicate checker or mandatory test/agent count.
  • Focused checks = only the tests, analyzer and lint for the touched files or package; full suites run once, in the final gate. A long run → HE-NO-FOREGROUND-POLL.
  • Direct test/build/check runs = limit ≈3× normal, ≥15 min (macOS: perl -e 'alarm shift; exec @ARGV' <s> <cmd>); hit → read log, kill leftover children, rerun.
  • Progress = same plan + remaining work. Retain the actual starting-baseline outcome + evidence; record later build results in Verification, in the commit that changes the proved code. No commit only records a result. Keep Status Ready and Verification Pending during an unblocked feature build; baseline repair follows HE Plan's Draft route above. A material decision or unavailable prerequisite → Draft + exact blocker/resume condition; preserve completed steps and continue independent authorized work. Never replace missing proof with a pass or N/A.

Parallel work + integration

  • Dispatch only independent complete behaviors with named owners, agreed interfaces and satisfied dependencies, within the task's delegation authorization. Small work stays with one builder. Check actual source and shared effects: generated files, lockfiles, fixtures, ports and databases can contend despite disjoint edit paths.
  • Use isolation when needed; serialize conflicting work and shared Git operations. In a shared checkout, one owner controls each write surface; workers avoid installs, builds or tests that mutate another worker's state. Worktrees do not isolate external resources.
  • Worker results = changed paths, actual commands/results and blockers in the normal handoff. Coordinator inspects the actual diff and shared contracts, integrates in dependency order, and runs required checks on the combined result. Preserve unowned changes. Worker passes or a clean merge do not prove integration.

Build complete → ready for ship

  • Reconcile the final diff against every accepted requirement and later correction, including the plan's explicit E2E disposition. Required local journeys must have Passed evidence; only deployment-dependent proof may remain Delivery under a Deploy target with its configured verifier. Confirmed review findings are resolved and no build blocker remains. Review claims against actual tests/runtime evidence; plan checkboxes alone are declarations.
  • Record actual focused-test/runtime evidence in the plan, check only verified acceptance items, set Status Complete + Verification Passed, then run the existing Complete check on the final integrated code. Ready checks own planning/baseline readiness; do not repeat them as an extra build gate. Failure → follow that gate's Draft/blocker recovery and resume the loop; never announce Ready for ship. Code/configuration changes after a pass need affected proof; batch them, then rerun the final gate once. Pre-push repeats that gate, so run no extra full check just before pushing.
  • After the gate passes, tell the user Ready for ship — local implementation and verification complete; delivery not performed. Handoff = implemented outcome + actual checks/runtime evidence + pending delivery requirements. The Complete plan already records the stage; do not rewrite it or repeat application checks solely to record this announcement. Authorized delivery continues through HE Ship; build alone adds no delivery authority. No BUILD.md or second state record.

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.