The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of processing external, potentially untrusted codebases.
Ingestion points: The agent is instructed to scan all files within a project directory using Read, Glob, and Grep tools (as detailed in references/analysis-checklist.md).
Boundary markers: The instructions lack specific delimiters or guardrails to prevent the agent from following instructions hidden within code comments or documentation of the analyzed system.
Capability inventory: The agent has access to Bash, Read, Grep, and Glob tools, which provides a broad capability set if an injection is successful.
Sanitization: No sanitization or filtering logic is mentioned for the content read from the target codebase.
[DATA_EXFILTRATION]: The skill instructions and checklists specifically guide the agent to locate and analyze sensitive configuration targets, including .env files, environment variables (os.environ), and authentication implementation details (JWT, Passport). While this is necessary for thorough reverse-engineering, it explicitly identifies high-value targets for data exposure within the project scope.

spec-miner