The Agent Skills Directory

[COMMAND_EXECUTION]: Uses git diff and git diff --cached to identify modified SKILL.md files in the repository.
[COMMAND_EXECUTION]: Executes grep commands to perform linting and consistency checks on skill files, such as verifying the presence of specific sections or phase naming conventions.
[DATA_EXFILTRATION]: Accesses local repository files including SKILL.md, CHANGELOG.md, README.md, and docs/workflow/model-routing.yml to read and update versioning metadata. All operations are local and no data is transmitted externally.
[PROMPT_INJECTION]: Ingests and processes content from user-modified files within the repository, which serves as an indirect prompt injection surface. Malicious data in these files could potentially affect the agent's logic during the documentation update process.
Ingestion points: git diff output and contents of skills/*/SKILL.md files.
Boundary markers: Absent.
Capability inventory: File modification, git command execution, and recursive grep searches.
Sanitization: Content from files is used without explicit validation or escaping.
[EXTERNAL_DOWNLOADS]: The documentation mentions an installation command npx skills add gtrabanco/agentic-workflow#claude. This refers to the official repository of the author and is provided as a setup instruction for users.

bump-skill