maintain-project-context
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation references a dependency on the
define-productskill from thetoy-crane/skillsrepository and provides instructions for manual installation vianpxif the skill is not already present. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external, potentially untrusted sources such as Pull Request (PR) bodies and remote issue trackers to synchronize repository documentation.
- Ingestion points: The skill reads merged PR bodies to find issue identifiers and fetches data from a remote issue tracker based on those IDs.
- Boundary markers: The instructions contain explicit safeguards, directing the agent to treat issues solely as coordination surfaces and not as authoritative sources for product intent or decisions. It also mandates that issue prose must not be imported into durable documentation files.
- Capability inventory: The skill is capable of reading and writing local project files (e.g.,
PRODUCT.md,GLOSSARY.md,docs/decisions/), performing Git operations to fetch branch data, and interacting with issue tracker APIs. - Sanitization: The skill uses specific text patterns for data extraction and limits its updates to sections of the issue tracker body explicitly marked by convention delimiters.
Audit Metadata