The Agent Skills Directory

[DATA_EXFILTRATION]: The skill explicitly instructs the agent to search for and document sensitive information, including environment variables, API keys, and database tables specifically named to contain secrets (e.g., *config*, *secret*, *setting*). While intended for auditing a project handoff, this automated hunting for credentials increases the risk of accidental exposure or leakage of system secrets.
[PROMPT_INJECTION]: The skill is highly vulnerable to Indirect Prompt Injection. It is designed to ingest and process a wide variety of untrusted external content, including project documentation, architecture diagrams, and source code repositories.
Ingestion points: Project documentation (Markdown/PDF), source code (all files), API responses, and database schemas/content (SKILL.md, Phase 1, 4, 5, 6).
Boundary markers: None identified. The instructions do not specify the use of delimiters or warnings to ignore instructions embedded in the analyzed data.
Capability inventory: The skill uses powerful tools including curl for network requests, git clone for code retrieval, and manus-mcp-cli for direct SQL execution on Supabase databases (SKILL.md, Phase 2, 5, 6).
Sanitization: None identified. The agent is encouraged to read and analyze implementation details directly.
[COMMAND_EXECUTION]: The workflow incorporates the use of command-line tools to interact with the environment, such as curl for probing API endpoints (Phase 5) and the GitHub CLI (gh auth status) for verifying repository access (GitHub Access Pattern section). This provides an execution path for instructions derived from potentially untrusted handoff data.

project-handoff-ingestion