cgagentharness-parity

Maintains CyClaw-to-harness parity documentation by classifying drift against harness code truth.

Updated Sep 7, 2026
One-click install
npx skills add https://github.com/cgfixit/CG-agent-harness --skill cgagentharness-parity-cgfixit
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cgagentharness-parity
Source: https://github.com/cgfixit/CG-agent-harness/tree/main/.claude/skills/cgagentharness-parity
Command: npx skills add https://github.com/cgfixit/CG-agent-harness --skill cgagentharness-parity-cgfixit

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Keeping cross-project parity documentation honest is hard: docs drift from code, and teams may be tempted to weaken security invariants just to match another product's behavior. This Skill provides a disciplined workflow for maintaining CyClaw↔CG-agent-harness parity contracts without ever weakening harness invariants. ## Core Features & Use Cases - Drift Classification: Tags every parity claim as FACT (observed in harness code, config, or tests) or INFERENCE (suspected intent or stale prose), so PR bodies stay evidence-based. - Status Reporting: Runs scripts/parity-status.py and produces a structured PASS/DRIFT/BLOCKED report summarizing drifts and invariant preservation. - Hard Refuse Rules: Explicitly blocks opening harness gates (agentic.enabled, allow_git_write_tools, api_key_optional) or loosening invariant_guard checks to chase parity. - Use Case: After a CyClaw merge, run this Skill to refresh docs/parity/* ledgers, record intentional divergences (e.g., fail-closed gates, Seatbelt isolation) as FACT, and defer CyClaw product-policy questions to the Advisor instead of mutating harness code. ## Quick Start Ask the AI to run the parity workflow to check docs/parity ledgers against current harness code and report any drift.

Frequently Asked Questions about cgagentharness-parity

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check parity between CyClaw and CG-agent-harness?▼

Read the docs/parity ledgers, then run python3 scripts/parity-status.py to compare documented claims against harness code. Classify each discrepancy as FACT (evidenced in code, config, or tests) or INFERENCE (unverified intent or stale prose).

How to classify documentation drift in a parity ledger?▼

Tag a claim as FACT when it is observed in harness code, assets/config.default.yaml, locking tests, or script output. Tag it as INFERENCE when it reflects suspected intent, stale prose, or CyClaw-side behavior not yet evidenced in the harness tree.

What is the truth order when parity docs conflict with code?▼

The authoritative order is code first, then assets/config.default.yaml, then INVARIANTS.md, then AGENTS.md, and parity docs last. When code moves, update the docs to match; never silently weaken code to match stronger doc claims.

Can harness security gates be opened to match CyClaw behavior?▼

No. The Skill hard-refuses opening agentic.enabled, allow_git_write_tools, or api_key_optional, and refuses loosening invariant_guard, jail, or hostile-argv asserts. Intentional divergences like fail-closed gates are recorded as FACT, not fixed.

What should I do if parity-status.py is missing from the checkout?▼

Report its absence as a FACT about the tree tip rather than inventing ledger rows. Fall back to INVARIANTS.md, AGENTS.md, and the relevant Rust modules, marking any proposed doc locations as INFERENCE only.

Does this Skill authorize pushing or merging parity changes?▼

No. The Skill only maintains and reports on parity documentation; it grants no authorization to push, merge, or arm gates, and performs no GitHub mutations on its own.