The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from repository files such as ADRs, design records, and source code which can serve as a vector for indirect prompt injection if those files contain instructions targeting the agent.
Ingestion points: Audits ADR files in docs/adr/, feature records in docs/fdr/, architecture documentation, and implementation files in src/.
Boundary markers: The instructions do not specify the use of delimiters or clear separation between the data being reviewed and instructions for the agent.
Capability inventory: The skill utilizes directory listing, file reading, and shell search commands (grep). It has the capability to write to the repository, although this is gated by user approval.
Sanitization: No specific content sanitization or validation logic is defined before processing the ingested file content.

adr-review