The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local validation script, scripts/validate_review.py, to verify findings.
Evidence: SKILL.md contains the instruction python3 scripts/validate_review.py findings.json.
The script performs schema and content validation on JSON data using standard libraries and does not involve network access or unsafe dynamic code evaluation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its primary function is to ingest and analyze untrusted data from external code repositories and pull requests.
Ingestion points: SKILL.md specifies that the agent should inspect diffs, code, tests, and documentation from pull requests and commits.
Boundary markers: There are no instructions for using specific delimiters or markers to isolate external data from the system prompt or agent instructions.
Capability inventory: The agent has the capability to execute local scripts and read repository files.
Sanitization: The scripts/validate_review.py script implements basic sanitization by enforcing ASCII text for titles and validating line ranges and field types in the generated findings.

harness-review