| name | execution-manifest |
| description | Use this skill when preparing work for a coding agent, remote agent, automation, CI workflow, or tool handoff. Creates a pinned execution manifest so agents consume explicit context instead of guessing latest sources, branches, documents, constraints, outputs, and verification commands. |
Execution Manifest
Use this skill to package a task for deterministic agent execution.
Principle
Agents consume manifests, not guesses. Resolve the latest intended inputs first, pin their versions, then execute against the pinned manifest.
Default Path
docs/00-context/execution-manifests/{manifest-id}.yaml
Use another path if the project already has a manifest convention.
Workflow
- Identify the execution goal.
- Resolve source references:
- repo URL or local path
- branch/ref
- commit SHA when available
- relevant document paths
- ticket/source references
- Pin inputs:
- approved docs
- ADRs
- constraints
- expected outputs
- Define permissions and boundaries:
- allowed writes
- forbidden files
- external actions needing approval
- Define verification:
- build/test commands
- quality gates
- expected evidence
- Write the manifest and point the agent/tool at it.
Manifest Template
Use references/execution-manifest-template.yaml.
Good Manifest Rules
- Prefer exact refs over "latest".
- Include every document the agent is expected to honor.
- Include output expectations, not just input context.
- Include rollback or stop conditions for risky work.
- Treat missing pins as a risk and call them out.