plan-0-v2-constitution

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its core function of processing untrusted external data into project documentation.
  • Ingestion points: The skill ingests untrusted data from the $ARGUMENTS variable as well as local files like README.md and CONTRIBUTING.md during the 'Context Gatherer' phase (Step 2).
  • Boundary markers: While the skill uses <!-- USER CONTENT START --> markers for content preservation during re-entrancy, it lacks explicit security delimiters or 'ignore' instructions to prevent the agent from obeying instructions embedded within the ingested files or arguments.
  • Capability inventory: The skill has significant file-write capabilities, allowing it to create and update multiple markdown files in the docs/project-rules/ and docs/domains/ directories (Steps 4, 5, 6, and 9).
  • Sanitization: The instructions do not define any sanitization, filtering, or validation steps for the content extracted from external sources before it is written to the project's authoritative doctrine files.
  • [SAFE]: The skill operates entirely within the local file system and does not perform any network operations, such as external downloads or data exfiltration.
  • [SAFE]: No hardcoded credentials, secrets, or sensitive system file paths (e.g., SSH keys, AWS configs) are accessed or exposed.
  • [SAFE]: The use of parallel subagents is an architectural pattern for processing speed and does not involve the execution of dynamic or obfuscated code.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 04:52 AM
Security Audit — agent-trust-hub — plan-0-v2-constitution