The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is primarily designed to execute the badger CLI tool using the command badger api review-context. This command uses local workspace paths and a temporary guidance file, representing standard tool integration.
[PROMPT_INJECTION]: The skill processes user conversation data to generate a 'guidance file' for the review tool, which is a surface for indirect prompt injection.
Ingestion points: Content from the current conversation and session work in SKILL.md.
Boundary markers: The instructions do not define specific delimiters for the interpolated content but mandate a concise, task-relevant summary.
Capability inventory: The shell command capability is limited to the badger utility.
Sanitization: The skill contains explicit negative constraints forbidding the inclusion of credentials, tokens, passwords, private keys, or .env values in the output.
[EXTERNAL_DOWNLOADS]: The skill provides a link to the badger installation documentation on GitHub (github.com/PVRLabs/aibadger). It strictly prohibits the agent from performing the installation or accessing the network itself, adhering to safety best practices.

badger-review