invariant-ace
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external artifact state and relevant files (e.g., source code, logs, and diffs) to identify software invariants. This creates an attack surface where malicious instructions embedded in these files could attempt to influence the agent's reasoning process.
- Ingestion points: The skill explicitly processes artifact_state_id (branch, diff_digest) and direction_state_id, along with relevant files for analysis as defined in SKILL.md and references/CODEX_SUBAGENTS.md.
- Boundary markers: There are no specific instructions for the agent to use delimiters or to ignore instructions embedded within the analyzed content.
- Capability inventory: The skill generates a Change Agenda and Invariant Ledger that drive subsequent implementation and review workflows.
- Sanitization: The skill lacks explicit sanitization or filtering logic for external inputs.
Audit Metadata