The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common developer tools through shell commands to perform context discovery and scope inventory. Commands used include git status, git rev-parse, rg, ls, cargo metadata, and make. These are expected operations for an auditing and review workflow performed on a local repository.
[DYNAMIC_EXECUTION]: The workflow includes the execution of a local Python script (plugins/imbue/scripts/citation_verifier.py) to verify that citations in the findings correspond to actual code in the repository. This represents a legitimate use of project-local tooling to automate quality checks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external repository data such as source code, configurations, and documentation, which creates a potential surface for indirect prompt injection if those files contain malicious instructions targeted at the agent.
Ingestion points: The skill instructions specify reading 'source files, configs, docs, specs' and capturing evidence from command outputs like git diff.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands when the agent is processing these external artifacts.
Capability inventory: The agent possesses the capability to execute shell commands and write to local files (e.g., .review/findings.json and docs/lessons-learned.md).
Sanitization: The workflow does not describe any sanitization or validation steps for the content of the ingested files before they are processed by the agent.

review-core