The Agent Skills Directory

DATA_EXFILTRATION (HIGH): The skill directs the agent to access highly sensitive files. Specifically, references/codebase-understanding.md instructs the agent to 'Analyze dotenv files and configuration'. Accessing .env files is a critical security risk as they frequently contain hardcoded API keys, database credentials, and other secrets.
EXTERNAL_DOWNLOADS (LOW): The skill facilitates the download and processing of code and data from arbitrary remote sources. Evidence in references/research-phase.md shows the use of repomix --remote <github-repo-url> to fetch and aggregate remote repositories and the docs-seeker skill to fetch documentation from external URLs. While intended for research, this behavior ingests untrusted data.
PROMPT_INJECTION (LOW): The skill is vulnerable to Indirect Prompt Injection due to its extensive ingestion of external, untrusted data. * Ingestion points: Remote GitHub repositories (repomix), GitHub PRs/Issues/Logs (gh command), and external documentation. * Boundary markers: None identified; there are no instructions to the agent to disregard instructions embedded within the research data. * Capability inventory: The agent can execute shell commands (repomix, gh), read/write local files, and spawn sub-agents (researcher, debugger). * Sanitization: None identified.

planning