The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted code and prose to perform cleanup tasks, creating a surface for potential indirect prompt injection attacks where malicious data could influence agent actions. * Ingestion points: Processes repository files including source code, READMEs, PR descriptions, and RFCs as specified in SKILL.md. * Boundary markers: The instructions lack explicit delimiters or safety warnings for the agent to ignore instructions embedded within the processed text. * Capability inventory: The skill is authorized to modify local files and execute shell commands such as git and bun. * Sanitization: There is no evidence of input sanitization or validation performed on ingested data before processing.
[COMMAND_EXECUTION]: The skill executes shell commands for scoping changes and verifying code integrity. * Evidence: Uses git merge-base and git diff for scoping changes, and bun run type-check, bunx tsc, and bun run test for verifying that edits do not break the build.

deslop