agentleFS
nishant-tamilselvannishant-tamilselvan/AEGIS/.github/skills/implementation-management/SKILL.md

implementation-management

Skill1 starsChanged 48 days agoMITView on GitHub

Plan, execute, resume and validate AEGIS phase-3 implementation from approved artifacts. Use when starting implementation, creating or transitioning WP-* work packages, recording IDEC-* decisions or evidence, reviewing target code, handling drift, or approving release.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    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:

    Follow enterprise-standards: retrieve current Approved Patterns, Software Delivery standards and related ADRs.

    line 25
  2. Very specific:

    Run readiness before any write: python -m artifact_tools implementation readiness <app-dir> <target-workspace> --standards-review verified

    line 27
  3. Distinctive:

    If blocked, do not initialize or edit code.

    line 29
  4. Distinctive:

    If ready, inspect target instructions and analogous code, ask Implementation Planner for the complete graph, then initialize:

    line 31
  5. Very specific:

    If the MCP is unavailable, state: "Enterprise Standards MCP unavailable — standards grounding skipped, manual review required".

    line 36
  6. Distinctive:

    every source row/document and source version listed by the package.

    line 45
  7. Distinctive:

    Never send an unbounded “implement the app” instruction.

    line 51
  8. Distinctive:

    Only packages in one explicit parallel group with disjoint target paths may be active concurrently.

    line 70
  9. Distinctive:

    all acceptance criteria and traced obligations are implemented.

    line 89
  10. Distinctive:

    only declared paths changed.

    line 90
  11. Distinctive:

    no secrets or unapproved scope are present.

    line 94
  12. Very specific:

    Run implementation status and implementation validate --strict before continuing.

    line 109

≈ “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

Implementation Management

Authoritative procedure for AEGIS phase 3. Approved requirements and architecture are executable specifications; target code is produced only through bounded work packages and actual verification evidence.

Ownership

  • Artifact Manager is the sole writer anywhere under docs/artifacts/<app>/, including implementation/implementation.md, decision.md and work-packages/.
  • ADR Author is the sole writer to architecture-decisions/.
  • Implementation code agents write only the active WP-* target paths in the separately resolved target workspace.
  • Implementation Reviewer is read-only.

Start procedure

  1. Resolve one docs/artifacts/<app> and an absolute target workspace.
  2. Follow enterprise-standards: retrieve current Approved Patterns, Software Delivery standards and related ADRs; check supersession and review health.
  3. Run readiness before any write: python -m artifact_tools implementation readiness <app-dir> <target-workspace> --standards-review verified
  4. If blocked, do not initialize or edit code. Route material gaps back through the architecture specialist, user HITL, Artifact Manager/ADR Author and Critic Reviewer.
  5. If ready, inspect target instructions and analogous code, ask Implementation Planner for the complete graph, then initialize: python -m artifact_tools implementation init <app-dir> <target-workspace> --target-branch <branch> --target-baseline <commit> --standards-review verified
  6. Artifact Manager creates proposed packages with implementation work-package new.

If the MCP is unavailable, state: "Enterprise Standards MCP unavailable — standards grounding skipped, manual review required". Initialize only with manual-review-required; strict validation and readiness remain blocked until a human records verified review.

Context packet for every package

Pass the implementer exactly one approved package plus:

  • canonical pointer and target baseline;
  • every source row/document and source version listed by the package;
  • accepted, non-superseded ADRs and current Enterprise Standards documents;
  • native interface contracts;
  • target instructions and analogous code;
  • dependencies' completion evidence.

Never send an unbounded “implement the app” instruction.

Work-package state machine

planned -> approved -> in-progress -> review -> complete

Allowed correction paths are in-progress -> blocked, review -> in-progress, and blocked -> approved|in-progress. Planned/approved work may be deferred or cancelled; deferred work may return to planned. Approval requires a named approver. Completion requires at least one evidence record and independent reviewer PASS.

Useful commands:

  • implementation work-package transition <app> <WP> approved --actor <actor> --approved-by <person>
  • implementation work-package transition <app> <WP> in-progress --actor <agent>
  • implementation work-package record-evidence <app> <WP> "<command>" pass --details "<result>"
  • implementation work-package transition <app> <WP> review --actor <agent> --review-status pass
  • implementation work-package transition <app> <WP> complete --actor artifact-manager

Only packages in one explicit parallel group with disjoint target paths may be active concurrently. See schema.

Decision policy

Record a tactical IDEC-* only when the choice is reversible, package-local, does not change approved behavior or controls, and does not create material risk. Record the question, choice, rationale, sources, paths and approver.

Pause and escalate when the choice changes scope, requirements, contracts, data ownership, security/privacy, deployment topology, residual risk, multiple components, or is hard to reverse. Obtain HITL, create/supersede an ADR, update affected artifacts, run strict review, and refresh package source versions. See decision escalation.

Definition of done

A package is complete only when:

  • all acceptance criteria and traced obligations are implemented;
  • only declared paths changed;
  • native format/lint/typecheck/build/test commands pass;
  • applicable contract, migration/rollback, security, accessibility, performance and observability checks pass;
  • no secrets or unapproved scope are present;
  • Implementation Reviewer returns PASS;
  • actual commands/outcomes are recorded as evidence.

The release gate additionally requires complete disposition of every active FR-*, NFR-*, BP-*, IF-*, DM-*, SEC-*, DEP-*, OBS-* and accepted ADR-*, strict validation, rollback readiness and explicit approval: python -m artifact_tools implementation release-approve <app> --approver <person>. Never deploy automatically.

Resume and drift

implementation.md is the sole resume pointer. On every resume compare recorded source versions, accepted ADR status, native contracts, Enterprise Standards currency and target baseline. Drift invalidates affected approvals and evidence until the planner reconciles the package. Run implementation status and implementation validate --strict before continuing.

Self-correction

Reviewer NEEDS-CHANGES returns the package to in-progress and the same implementer. Material findings follow the architecture remediation protocol, not an in-code workaround. Artifact Manager updates state only after the orchestrator accepts observed evidence.

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.