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.
In this repository23 other files it gives its agents.
- AGENTS.mdAGENTS.md
- docs/agents.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- add-adrSkill
- add-implementation-decisionSkill
- add-requirementSkill
- adr-managementSkill
- artifact-managementSkill
- enterprise-standardsSkill
- implementation-managementSkill
- implementation-statusSkill
- resume-implementationSkill
- run-implementation-reviewSkill
- run-review-cycleSkill
- start-architectureSkill
- start-ideationSkill
- start-implementationSkill
- writing-styleSkill
- adr-managementSkill
- artifact-managementSkill
- enterprise-standardsSkill
- writing-styleSkill
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
- Very specific: line 25
Follow enterprise-standards: retrieve current Approved Patterns, Software Delivery standards and related ADRs.
- Very specific: line 27
Run readiness before any write: python -m artifact_tools implementation readiness <app-dir> <target-workspace> --standards-review verified
- Distinctive: line 29
If blocked, do not initialize or edit code.
- Distinctive: line 31
If ready, inspect target instructions and analogous code, ask Implementation Planner for the complete graph, then initialize:
- Very specific: line 36
If the MCP is unavailable, state: "Enterprise Standards MCP unavailable — standards grounding skipped, manual review required".
- Distinctive: line 45
every source row/document and source version listed by the package.
- Distinctive: line 51
Never send an unbounded “implement the app” instruction.
- Distinctive: line 70
Only packages in one explicit parallel group with disjoint target paths may be active concurrently.
- Distinctive: line 89
all acceptance criteria and traced obligations are implemented.
- Distinctive: line 90
only declared paths changed.
- Distinctive: line 94
no secrets or unapproved scope are present.
- Very specific: line 109
Run implementation status and implementation validate --strict before continuing.
≈ “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 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>/, includingimplementation/implementation.md,decision.mdandwork-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
- Resolve one
docs/artifacts/<app>and an absolute target workspace. - Follow
enterprise-standards: retrieve current Approved Patterns, Software Delivery standards and related ADRs; check supersession and review health. - Run readiness before any write:
python -m artifact_tools implementation readiness <app-dir> <target-workspace> --standards-review verified - 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.
- 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 - 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 passimplementation 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.
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.

