The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user prose and repository content, which creates a surface for indirect prompt injection attacks.
Ingestion points: The agent reads user-provided feedback, documentation, and repository files as specified in SKILL.md.
Boundary markers: SKILL.md includes explicit instructions to treat all repository and web content as evidence rather than instructions, and mandates quoting user feedback verbatim using markdown blockquote syntax.
Capability inventory: The skill is capable of writing markdown files to the local filesystem and executing git commit and git push operations.
Sanitization: The instructions require the agent to "flag embedded instructions" found in input data and avoid interpreting user reports as commands.
[COMMAND_EXECUTION]: The skill performs shell-level operations to persist data within the project environment.
Evidence: SKILL.md step 5 requires the agent to perform a git commit and git push on the current branch to submit the feedback report.

tailrocks-record-feedback