The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute git diff and a local script (scripts/review-ai-residuals.sh) to collect data for analysis. This is consistent with its stated purpose of performing code reviews and scanning for artifacts like 'AI residuals' or hardcoded secrets.
[PROMPT_INJECTION]: The skill represents an indirect prompt injection surface as it ingests and processes untrusted data from external sources.
Ingestion points: Reads code changes via git diff and project tasks via Plans.md.
Boundary markers: The instructions define a structured JSON output format for the agent, which helps in separating data from instructions, but no explicit 'ignore embedded instructions' markers are used for the input data.
Capability inventory: The skill has access to shell execution (Bash), task management (Task), and file system tools (Read, Grep, Glob, Edit).
Sanitization: There is no evidence of sanitization or escaping of the content being reviewed before it is processed by the AI model.

harness-review