The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run a local Python script (scripts/score.py) for processing evidence and calculating scores. The script relies solely on standard library modules and performs operations consistent with the skill's stated purpose.- [DATA_EXFILTRATION]: Evaluation results are persisted locally to a ./_judge/ directory or a location defined by the JUDGE_DIR environment variable. No network operations or attempts to access sensitive system files were detected.- [PROMPT_INJECTION]: The skill processes untrusted user content (drafts, plans, ideas), representing an indirect prompt injection surface. This is a low-risk concern given the skill's specific task of evaluating text rather than executing it.
Ingestion points: User-provided content such as LinkedIn posts, specs, and project plans enter the context for analysis as described in SKILL.md.
Boundary markers: The skill does not define explicit delimiters or 'ignore' instructions to isolate user content from the feedback logic.
Capability inventory: The skill has the capability to write JSON files to the local file system via the scripts/score.py utility.
Sanitization: SKILL.md provides instructions to derive file 'slugs' using kebab-case and no special characters, which helps prevent path traversal or filename manipulation when saving results.

what-works-feedback-judge