The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes source code files which are externally provided and untrusted. Maliciously crafted code could attempt to influence the agent's summary or recommendations through specific AST patterns or comments that match the skill's detectors.
Ingestion points: Source code files read from the local file system (SKILL.md).
Boundary markers: None identified in the provided modules for the analyzed source content.
Capability inventory: File system read access, shell command execution (via documentation for verification and visualization), and generation of structured review findings.
Sanitization: No explicit sanitization of AST-extracted text (e.g., variable names or anchors) is mentioned before inclusion in reports.
[EXTERNAL_DOWNLOADS]: The modules/kuva-visualization.md file provides instructions to download and install kuva, a Rust scientific plotting library, using cargo install kuva from the official Crates.io package registry. This is documented for the purpose of visualizing benchmark results.
[COMMAND_EXECUTION]: The skill's verification workflow includes executing a local script plugins/imbue/scripts/citation_verifier.py to validate that detected hotspots exist in the source code. It also includes documentation for executing kuva, pytest, and git commands as part of the performance analysis process.

performance-review