AI Agent Skills agent-readable instructions
The public BRIK64 skill helps agents shape requirements into PCD workflows without silent managed-file writes.
- Requirement shaping: Explain inputs, outputs, external assumptions, and fail-closed behavior.
- Monomer selection: Choose from the catalog instead of inventing unsupported operations.
- PCD authoring: Keep branches, returns, and EVA structure explicit and reviewable.
- Evidence reporting: Report commands run, files changed, evidence status, and remaining assumptions.
- Boundary: the skill is guidance for agents. It is not a certificate for the repository or generated code.
AI Agent Skills
Skills for agents that need to work inside bounds.
The public BRIK64 skill helps agents shape requirements into PCD workflows without silent managed-file writes.
Agent Workflow
The skill gives agents a bounded way to participate in BRIK64 work.
It keeps generated work aligned with the public docs, PCD rules, and explicit user approval for managed instructions.
What The Skill Guides
The public docs define the behavior agents should follow during BRIK64 work.
Requirement shaping
Explain inputs, outputs, external assumptions, and fail-closed behavior.
Monomer selection
Choose from the catalog instead of inventing unsupported operations.
PCD authoring
Keep branches, returns, and EVA structure explicit and reviewable.
Evidence reporting
Report commands run, files changed, evidence status, and remaining assumptions.
Managed Files
Managed-section changes must be explicit, reviewable, and approved by the user.
Preview first
Show the managed instruction block before any write.
Use markers
Respect public managed-section marker contracts when supported.
Report conflicts
Stop and explain when local instructions conflict with public guidance.
Make agent work readable before it becomes code.
Use the public skill and docs to keep PCD structure, .brik traceability, and evidence status visible.