The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes research goals and paper summaries derived from external sources.
Ingestion points: Reads from reaper-workspace/notes/literature.md and reaper-workspace/notes/paper-summary.md, which contain external research content.
Boundary markers: None identified; instructions do not explicitly warn the agent to ignore instructions embedded in the processed research text.
Capability inventory: The skill performs file read operations across various directories and file write operations to problem-statement.md and ideas.md.
Sanitization: No validation or sanitization of the external research data is performed before processing.
[COMMAND_EXECUTION]: The skill instructs the agent to perform filesystem discovery by resolving absolute paths for a sibling skill ('reaper'). It lists several common installation directories in the user's home folder (e.g., ~/.claude/skills/, ~/.agents/skills/) to facilitate this. While this involves probing the host environment, it is used locally to resolve configuration references (model.md, impossibility-results.md) within the same vendor's ecosystem.

formalize-problem