apiVersion: devrunbook.io/v1alpha1
kind: Playbook
metadata:
  id: repository-understanding.agents-instructions
  slug: agents-instructions
  version: 1.0.0
  title: Generate Repository AGENTS.md Guidance
  summary: Create reviewed persistent Codex instructions from real repository commands, protected paths and engineering policies.
  category: repository-understanding
  tags:
  - codex
  - agents.md
  - governance
  lifecycle: reviewed
  riskTier: moderate
  authors:
  - name: DevRunbook Core Team
  license: MIT
package:
  files:
  - path: prompt.md
    role: template
    digest: true
    exportByDefault: false
  - path: README.md
    role: documentation
    digest: true
    exportByDefault: false
  - path: CHANGELOG.md
    role: changelog
    digest: true
    exportByDefault: false
  - path: examples/minimal.yaml
    role: example
    digest: true
    exportByDefault: false
  - path: evaluations/static-structure.yaml
    role: evaluation
    digest: true
    exportByDefault: false
spec:
  type: guided
  intent:
    problem: Development work around generate repository agents.md guidance is often underspecified, inconsistently executed
      or reported without enough evidence.
    outcome: Create reviewed persistent Codex instructions from real repository commands, protected paths and engineering
      policies.
    whenToUse:
    - Use this playbook when the repository needs a bounded generate repository agents.md guidance task with explicit evidence
      and completion criteria.
    - Use it when Codex should follow a repeatable workflow rather than improvise from a one-line request.
    whenNotToUse:
    - Do not use it when the desired outcome or authority boundaries are still materially undecided.
    - Do not use it to access unavailable production credentials, bypass safeguards or claim validation that cannot be performed.
  modes:
  - plan
  defaultMode: plan
  autonomy:
    min: diagnose
    max: plan
    default: plan
  inputs:
  - key: instructionScope
    label: Instruction scope
    description: Choose where durable Codex instructions should apply.
    type: enum
    required: true
    sensitive: false
    includeInOutput: true
    default: layered
    options:
    - global
    - repository
    - directory-specific
    - layered
  - key: directoryOverrides
    label: Directory overrides
    description: List directories that need stricter or different instructions and explain why.
    type: key-value-list
    required: false
    sensitive: false
    includeInOutput: true
    default: []
  compatibility:
    repositoryRequired: true
    languages: []
    frameworks: []
    packageManagers: []
    databases: []
    deploymentTypes: []
    requiredProfileCapabilities: []
    incompatibleConditions: []
  guardrails:
  - id: guardrail-1
    severity: blocking
    text: Keep durable repository rules separate from the current one-time task.
  - id: guardrail-2
    severity: blocking
    text: Never place secrets, private tokens or machine-specific absolute paths in AGENTS.md.
  - id: guardrail-3
    severity: blocking
    text: Do not claim a command is mandatory unless repository evidence or an explicit policy supports it.
  workflow:
  - id: inventory-existing
    title: Inventory existing instructions
    instruction: Read all applicable AGENTS.md and override files and determine their effective hierarchy.
    required: true
  - id: collect-rules
    title: Collect durable rules
    instruction: Extract verified commands, protected paths, architecture boundaries, testing expectations and Git policies.
    required: true
  - id: separate-scopes
    title: Separate scopes
    instruction: Assign global, repository and directory-specific rules to the narrowest correct location.
    required: true
  - id: draft-files
    title: Draft instruction files
    instruction: Produce complete suggested files without overwriting existing instructions.
    required: true
  - id: check-conflicts
    title: Check conflicts
    instruction: Identify contradictory rules, duplicate guidance and unsafe instructions before finalizing.
    required: true
  - id: handoff-review
    title: Prepare review notes
    instruction: Explain every material rule, its evidence and where human confirmation is still required.
    required: true
  validation:
    commandRoles: []
    checks:
    - id: check-1
      type: assertion
      description: Suggested instructions contain only durable, evidenced rules.
      blocking: true
      evidence: Referenced files, command results or explicit review notes.
    - id: check-2
      type: assertion
      description: The hierarchy and all conflicts or overrides are explicit.
      blocking: true
      evidence: Referenced files, command results or explicit review notes.
  completion:
    criteria:
    - Durable rules are separated from one-time task instructions.
    - Suggested hierarchy and review notes are included.
    - Validation evidence and unresolved limitations are reported honestly.
  failurePolicy:
    onValidationFailure: Investigate failures caused by the current work, repair them when they remain within scope, rerun
      affected validation and report any genuine blocker without claiming success.
    onAmbiguity: Use repository evidence and existing conventions for minor reversible choices. Preserve current behavior
      and stop before any material irreversible decision that the specification does not resolve.
    onMissingContext: Inspect the repository for missing non-sensitive context. Never invent commands, credentials, production
      behavior or validation results; report what remains unavailable.
    onOutOfScopeCause: Explain the evidenced out-of-scope cause, avoid unrelated changes and provide the smallest safe follow-up
      recommendation.
    onExternalDependencyUnavailable: Use an approved local substitute or fixture only when it preserves the behavior under
      test. Otherwise record the blocked validation and do not claim the external path succeeded.
    onUnableToReproduce: Record attempted reproduction, environment and observed evidence. Do not apply speculative production
      changes; provide the narrowest next diagnostic action.
  reporting:
    sections:
    - id: outcome
      title: Outcome
      required: true
      description: State the delivered result or audit conclusion without overstating evidence.
    - id: evidence
      title: Evidence and scope
      required: true
      description: List inspected or changed areas and the evidence supporting the result.
    - id: validation
      title: Validation
      required: true
      description: Report commands, manual checks and their actual outcomes.
    - id: risks
      title: Risks and limitations
      required: true
      description: State residual risk, inaccessible evidence and untested conditions.
    - id: follow-up
      title: Recommended follow-up
      required: true
      description: List the smallest useful next actions or state None.
  template:
    main: prompt.md
    partials: []
  exports:
    prompt: true
    markdown: true
    runPack: false
    agentsSuggestion: true
quality:
  reviewStatus: editorial-reviewed
  testedStacks: []
  knownLimitations:
  - Repository-specific effectiveness depends on the accuracy of the selected profile and the evidence available to Codex.
  evaluationCaseIds:
  - agents-instructions.static-structure