The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which could potentially contain malicious instructions aimed at influencing the review outcome.
Ingestion points: Data enters the agent context via git diff, git log, journey evidence, and GitHub issues through gh issue view as specified in SKILL.md.
Boundary markers: The instructions define separate prompts for sub-agents with specific briefs, but there are no explicit delimiters or instructions to ignore embedded commands within the ingested text.
Capability inventory: The skill utilizes git for repository inspection and gh for issue retrieval; it does not explicitly perform file-write or arbitrary network operations beyond these tools.
Sanitization: There is no evidence of sanitization or escaping of ingested data before it is passed to the sub-agent prompts.
[COMMAND_EXECUTION]: The skill instructions rely on the execution of shell commands to retrieve information for the review process.
Evidence: SKILL.md specifies the use of git diff, git log, git rev-parse, gh issue view, and gh issue list to gather change context and requirement specifications.

code-review