midi-backplane.git · FCIS_AUDIT.md

revision 23efa75425f4663d06e8d1a64c85628a880a0462 · complete file
# FCIS Audit (RFC 0001)

## Executive Summary
Status: Compliant.
Top findings:
- FCIS layers are present: `AGENTS.md`, `PLANS.md`, and `.codex/skills/repo-tooling/SKILL.md`.
- Procedural command lists are centralized in the Skills layer and referenced from docs.
- No MCP workflow/policy in repo (PASS).

## Repository Inventory
- `AGENTS.md`: behavioral invariants and routing.
- `PLANS.md`: long-task intent protocol template.
- `.codex/skills/repo-tooling/SKILL.md`: command recipes for lint/facts/tests.
- `README.md`: overview; points to skill for commands.
- `tools/midi-schema-lint/README.md`, `tools/midi-schema-to-facts/README.md`: tool summaries; point to skill for usage.

## Compliance Matrix
| Requirement ID | Requirement | Status | Evidence | Fix Recommendation |
| --- | --- | --- | --- | --- |
| FCIS-AGENTS-1 | Root `AGENTS.md` exists and defines invariants/routing | PASS | `AGENTS.md`: "## Invariants", "## Routing" | None |
| FCIS-AGENTS-2 | `AGENTS.md` excludes procedures/tool config | PASS | `AGENTS.md`: no command lists or step-by-step instructions | None |
| FCIS-PLANS-1 | `PLANS.md` exists and is required by AGENTS | PASS | `AGENTS.md`: "Use `PLANS.md`..."; `PLANS.md` present | None |
| FCIS-PLANS-2 | PLANS is intent protocol (no procedures) | PASS | `PLANS.md`: template fields for intent/risks/acceptance | None |
| FCIS-SKILLS-1 | Skills exist for procedures with required sections | PASS | `.codex/skills/repo-tooling/SKILL.md`: Purpose/When/Steps/Output | None |
| FCIS-ORTHO-1 | No responsibility overlap across layers | PASS | `README.md` and tool READMEs point to skill; AGENTS/PLANS declarative | None |
| FCIS-MCP-1 | MCP not required; no MCP workflow/policy in repo | PASS | No MCP references found | None |

## Orthogonality Violations
None observed after remediation. Procedural command lists now live in `.codex/skills/repo-tooling/SKILL.md` and are referenced from docs.

## Risks / Drift Vectors
- Procedural guidance could drift back into README/tool docs; keep command lists in skills.
- Plan protocol adoption relies on contributors following `AGENTS.md` routing.