The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive repository files including .env and internal git state files. Specifically, catalog/rebase-worktrees.md reads .env files to compare and synchronize keys across worktrees, and catalog/resolve-conflict.md reads internal git files such as .git/rebase-merge/onto to analyze conflict context. Note that no network exfiltration was detected.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository history and file contents, which could be used to influence the agent's behavior during complex operations like conflict resolution.\n
Ingestion points: catalog/resolve-conflict.md reads commit messages via git log and parses conflict markers in file contents; catalog/branch-recommend.md analyzes commit history.\n
Boundary markers: No specific delimiters or instructions to ignore embedded content are used when reading commit messages or file diffs.\n
Capability inventory: The skill has significant local capabilities including running package manager installation commands (npm install, bun install), executing project scripts from package.json, and deleting branches or worktrees (git branch -D, git worktree remove).\n
Sanitization: There are no explicit instructions for the agent to sanitize or escape data read from the repository before processing it.\n- [COMMAND_EXECUTION]: The skill executes various shell commands for git operations and project maintenance. This includes destructive operations like force-deleting branches in catalog/branch-clear-gone.md and running arbitrary scripts defined in the repository's configuration (e.g., linting and typechecking in catalog/commit.md).

git-lgnh