concept-guard-cn
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements an automated context injection mechanism that reads project specification files to provide instructions and boundaries to the agent.\n
- Ingestion points: Local specification files such as CONCEPT.md, PIPELINE.md, and SYNCS.md are read by the shell scripts drift-context.sh and post-check.sh located in the runtime/scripts directory.\n
- Boundary markers: Extracted content is presented to the agent using structural headers (e.g., [filename ## section]) and is accompanied by instructional prompts that direct the agent to verify changes against these declared boundaries.\n
- Capability inventory: The skill is configured for file system operations including reading, writing, and searching, and it utilizes shell execution capabilities through the integrated Claude Code hooks system.\n
- Sanitization: The extraction scripts use pattern matching with sed to isolate specific markdown sections, but they do not perform additional sanitization or filtering of the text content before it is injected into the agent's context.
Audit Metadata