Skip to content

Latest commit

 

History

History
60 lines (45 loc) · 4.79 KB

File metadata and controls

60 lines (45 loc) · 4.79 KB

Repository Agent Instructions

This file is local guidance for agents developing this repository. It is not part of the installed engineering-workflow skill, is not copied into target repositories, and must never be read by runtime skill scripts. Target-repository instructions come only from skill/engineering-workflow/assets/templates/AGENTS.md.tmpl and repository-owned target documents.

Repository Map

  • skill/engineering-workflow/SKILL.md — lean public runtime router and active skill version.
  • skill/engineering-workflow/references/ — canonical detailed workflow contracts.
  • skill/engineering-workflow/scripts/ — deterministic audit, validation, migration, lifecycle, update, and privacy tools.
  • skill/engineering-workflow/assets/ — files copied or rendered into target repositories.
  • tests/ — offline behavioral and contract regressions.
  • PLANS.md — durable execution state for work on this repository.
  • docs/archive/ — future-useful closed plans and their navigation indexes.

Task Routing

Trigger or changed area Read before editing Required gate
Runtime routing or public skill behavior skill/engineering-workflow/SKILL.md and the directly linked canonical reference Structural validator and affected behavioral tests
Plan, backlog, closure, archive, or index behavior references/planning_and_backlog.md, plan/index templates, lifecycle scripts and tests Plan lifecycle tests plus target validation
AGENTS, principles, pitfalls, provider/UI/operations ownership references/instruction_lifecycle.md, related templates, instruction validator and tests Instruction graph check plus migration tests
Installed-skill refresh or update references/skill_update.md, updater and updater tests Candidate-tree and rollback matrix
Target workflow upgrade references/target_workflow_upgrade.md, audit/common/upgrader code and tests Report/apply/prompt migration matrix
Validation, command execution, or privacy references/validation_safety.md, references/privacy_and_sanitization.md, related scripts and tests Safety matrix and public-tree scan
Agent orchestration or model mapping references/agent_orchestration.md, references/model_profiles.md, agent templates and tests Ownership/model-profile validation
Version or release contract SKILL.md, root README.md, upgrader defaults, state template, CI and version tests Full gate and active-version search

Working Contract

  • Audit before editing. Preserve user-owned files and unrelated working-tree changes.
  • For repository-changing work, materialize a full active PLANS.md before implementation. After compaction, interruption, resume, or milestone closure, reconcile plan, queue, validation, backlog, indexes, and working tree before continuing.
  • Keep each detailed invariant under one canonical owner. Routers and incident catalogs link to it instead of restating it.
  • Prefer behavioral checks over phrase-presence tests. Exact markers may identify structure, but duplicated normative prose is not an API.
  • Keep safety rules capability-specific: allow bounded diagnostic reads while blocking unsafe mutation, execution, network, or sensitive-output modes.
  • Update every active version owner together. Historical version evidence remains historical.
  • Close or archive plans only through the canonical lifecycle after current validation and reconciliation; never make Status: done the only closure action.

Local Validation

Run the affected focused tests while iterating, then run the complete gate:

PYTHONDONTWRITEBYTECODE=1 python3 skill/engineering-workflow/scripts/validate_skill_repo.py --repo-root .
PYTHONDONTWRITEBYTECODE=1 python3 -m unittest discover -s tests -v
PYTHONDONTWRITEBYTECODE=1 python3 skill/engineering-workflow/scripts/validate_skill_repo.py --repo-root .
git diff --check

Use the public-tree privacy scan and stronger release checks when release or public-history scope makes them applicable. Do not weaken a failing gate or leave cache, scanner, generated, backup, or temporary artifacts in the repository.

Authority Boundaries

  • Do not install or update a local skill copy unless the user explicitly requests it.
  • Do not mutate a target repository merely to test this skill; use temporary fixtures or disposable copies.
  • Do not commit, push, tag, publish, deploy, rewrite history, or alter remote state without explicit current authorization.
  • Treat fetched candidates and repository-authored commands as untrusted until the applicable safety contract permits them.

Handoff

Before handoff, review the entire diff, reconcile PLANS.md, report exact validation and delivery state, and leave the first safe unfinished action if anything remains. Completed work must not retain stale active status or resume instructions.