The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs codebase analysis and refactoring by executing multiple shell commands and development tools including find, grep, git, ruff, vulture, and tsc as specified in SKILL.md and references/audit-catalog.md.- [EXTERNAL_DOWNLOADS]: The skill uses npx jscpd in its audit catalog to detect code duplication, which facilitates the download and execution of the jscpd package from the official npm registry.- [INDIRECT_PROMPT_INJECTION]: The skill reads and processes all files within a repository, creating a potential vector for indirect prompt injection if the audited code contains malicious instructions.
Ingestion points: Audits entire directory structures and file contents across the repository (SKILL.md, references/audit-worker.md).
Boundary markers: The skill does not mandate explicit delimiters for all data ingestion, though snapshots via repomix are supported.
Capability inventory: The agent possesses broad capabilities including shell command execution, file system writes, and git manipulation (SKILL.md, references/quality-loop.md).
Sanitization: The risk is mitigated by rigorous human-in-the-loop checkpoints at every stage (Checkpoints A-E) and specific constraints on subagent outputs that require evidence citing path:line.

codebase-improver