The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is programmed to locate and read sensitive configuration files and environment variables to document the system's setup.
Evidence: Exploration patterns in SKILL.md target sensitive environment strings like os.environ and config settings.
Evidence: The reference guides references/analysis-checklist.md and references/analysis-process.md explicitly instruct the agent to find and read .env files.
[PROMPT_INJECTION]: The skill has an inherent attack surface for indirect prompt injection as it processes untrusted source code from external projects.
Ingestion points: Code content is read using Read, Grep, and Glob tools as defined in the analysis workflow.
Boundary markers: The instructions do not include specific delimiters or warnings to ignore malicious instructions embedded in the source code.
Capability inventory: The skill has access to the Bash, Read, Grep, and Glob tools and can write files to the local filesystem.
Sanitization: No content filtering or sanitization process is described for the data read from the codebase before it is processed by the agent.

spec-miner