The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code content from git branches to identify and remove 'slop', creating a vulnerability to instructions embedded within the analyzed data.
Ingestion points: File content is ingested through git diff, git show main:<filepath>, and direct file reads during the comparison step in SKILL.md.
Boundary markers: Absent. There are no delimiters or specific instructions to ignore potential commands embedded within the source code or comments being reviewed.
Capability inventory: The skill possesses file-reading capabilities via git and file-modification capabilities via the Edit tool.
Sanitization: Absent. The skill does not perform validation, escaping, or filtering of the code content before the agent processes it.
The broad instruction to 'Carefully fix anything you uncover' in Step 3 increases the risk that an agent might follow malicious instructions found in the code.

deslop