The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references its documentation and source code hosted on GitHub (github.com/aipoch/medical-research-skills). This is a reference to a well-known service and the author's own repository, which is considered safe behavior.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of reviewer comments and editor decision letters.
Ingestion points: Inputs provided by the user in response to Step 1 and Step 2 in SKILL.md (Reviewer comments, editor letters, manuscript weaknesses).
Boundary markers: None identified in the prompt instructions to delimit external content from instructions.
Capability inventory: None. The skill does not use any tools for code execution, file writing, or network communication; it only produces structured text output.
Sanitization: None. The skill processes the text as provided by the user.
Risk Assessment: Given the total lack of destructive capabilities or external side effects, this surface represents a minimal security risk.
[METADATA_POISONING]: The eval_report_revision-strategy-planner_result.json file appears to be a pre-generated evaluation report for the skill. While it contains self-claims about its own safety ('security: PASS'), these have been disregarded in favor of independent analysis. The metadata in SKILL.md is accurate and not deceptive.

revision-strategy-planner