phionyx-compliance

Evidence-grade compliance report drafts from Phionyx RGE v0.2 audit chains. AGPL-3.0 · Python 3.10+ · alpha (v0.7.0 cycle)

phionyx-compliance turns a signed envelope chain produced by phionyx-core, phionyx-mcp-server, or any other Phionyx component into a framework-shaped markdown draft report — suitable as input to a compliance officer or auditor's review.

It does not certify. Reports are evidence-oriented mappings; a lawyer or auditor must review before any use that implies legal posture.

Sixty-second usage (target spec; lands at v0.7.0)

pip install phionyx-compliance
phionyx-compliance generate \
  --trace trace-e2dd588aaf4d4c97 \
  --template eu-ai-act-article-13 \
  --out reports/<DATE>-article-13-draft.md

What it does:

1. Reads the envelope chain at ~/.phionyx/mcp_audit/<trace_id>/ (or the path the operator passes via --chain-root).
2. Walks the chain and extracts the per-envelope evidence the chosen template requires.
3. Renders a markdown draft against the template's prompts and the chain's actual evidence.
4. Includes the canonical disclaimer at the head and tail of the document.

Supported templates (v0.7.0 ship list)

Each template is a self-contained directory under templates/<framework>/:

templates/
└── eu-ai-act-article-13/
    ├── template.md          (markdown with {{placeholders}})
    ├── schema.json          (Pydantic JSON Schema for the data inputs)
    ├── mapping.yaml         (how envelope fields → template placeholders)
    └── README.md            (framework-specific notes + disclaimer text)

Composition with the Phionyx stack

phionyx-compliance is read-only over the envelope chain. It does not modify chain state, does not re-sign anything, and does not produce envelopes of its own. The output is a markdown file plus an optional JSON-formatted summary (--format=json).

This composes cleanly with:

• phionyx-pipeline-mcp — self-claim gate envelopes feed the "agent's own attestations" section of every framework template.
• phionyx-mcp-server — third-party tool-call envelopes feed the "tool-call audit" section.
• phionyx-eval-inspect — Inspect AI .eval exports include the same envelope chain; the compliance draft can cite the .eval log id as the reviewer-runnable evidence pointer.

Plugin command (lands at v0.7.0 with F12++)

The phionyx-claude-code-plugin will gain /phionyx:evidence-report that shells out to this package and renders the draft inline in the Claude Code chat. Pattern preserves the v0.5.1 plugin extraction discipline (no Anthropic plugin content copied verbatim).

Status (2026-05-26)

• W1.1 — scaffold landed. Package skeleton, version pin to phionyx-core>=0.6.0, CLI entry-point declared.
• W1.2 — template substrate landed. First template: eu-ai-act-article-13 v1.0.0. Renderer + CLI subcommands (list-templates / describe / render-sample / generate --sample). 15/15 tests.
• W1.3 — chain → inputs walker landed. ChainView.from_disk() + resolve_inputs() (4-category mapping resolver). Real generate --trace <id> works. 25/25 tests.
• W1.4 — plugin integration landed. /phionyx:evidence-report slash command in the Claude Code plugin wraps the CLI.
• W1.5 — additional templates landed (this commit). Three new framework templates at v1.0.0 each:
  • nist-ai-rmf-1 (NIST AI RMF 1.0 — MAP/MEASURE/MANAGE/GOVERN)
  • iso-iec-42001 (ISO/IEC 42001:2023 — AI Management System)
  • owasp-agentic-ai-v1 (OWASP Agentic AI Threats v1.0 — threat coverage) Cross-template parity test suite (test_all_templates.py, 5 tests). 30/30 total tests pass.

v0.7.0 W1 complete. W2 (F4 reasoning audit + F8 RAG audit) is the next milestone.

See docs/DESIGN.md for design notes.

Disclaimer (canonical)

Reports produced by this package are evidence-oriented mappings, not legal compliance guarantees. A lawyer or auditor must review before any use that implies legal posture. Phionyx does not certify AI systems; it produces the evidence chains that certifications rely on.

License

AGPL-3.0-or-later. See LICENSE.

Citing

If you use phionyx-compliance in academic or policy work, cite the parent project: Abak, A. T. (2026). Phionyx Research — Runtime Evidence Layer for Agentic AI. ORCID 0009-0002-3718-4010.