The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data (user-provided TypeScript code and git diffs) which could contain malicious instructions intended to manipulate the agent's behavior during the review or refactoring process.
Ingestion points: The workflow in references/review-and-fix-changes.md involves reading file contents and diffs via git status, git diff, and git show.
Boundary markers: The skill uses git logic to define a clear review scope, which provides a logical boundary, but it lacks explicit instructions to ignore or sanitize embedded instructions within the code being analyzed.
Capability inventory: The skill grants the agent the ability to execute shell commands (git, jscpd) and perform file modifications (applying 'obvious fixes').
Sanitization: No specific content sanitization or validation of the ingested code for prompt injection patterns is implemented.
[COMMAND_EXECUTION]: The skill uses several shell commands for its core functionality.
In SKILL.md, it executes git rev-parse --show-toplevel and bunx --yes jscpd to perform code duplication analysis.
In references/review-and-fix-changes.md, it uses a variety of git commands (status, diff, show, log, merge-base) to determine the scope of changes to be reviewed. These are standard development tools and their use is consistent with the skill's stated purpose.
[EXTERNAL_DOWNLOADS]: The skill utilizes bunx --yes jscpd in SKILL.md. This command downloads the jscpd package from the npm registry and executes it. This is a well-known code analysis utility, and the download is performed from a standard service (npm).

typescript-best-practices