The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to perform repository discovery and record metadata.
Git commands: Uses git status, git rev-parse, git log, git diff, git show, and git blame to analyze the repository state and evolution.
Search tools: Employs rg (ripgrep) to inventory repository files while excluding build and configuration directories.
File generation: Uses a shell heredoc with cat to write metadata into openwiki/.last-update.json, involving shell interpolation for dates and git hashes.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because its primary function is to ingest and summarize external data from the repository.
Ingestion points: Reads repository trees, configuration files, READMEs, entrypoints, and domain-specific source files across the entire codebase.
Boundary markers: The instructions do not specify using XML tags or other delimiters to isolate untrusted file content during the summarization process.
Capability inventory: The skill possesses the capability to execute shell commands and modify specific root-level instruction files (AGENTS.md and CLAUDE.md) in addition to creating the documentation tree.
Sanitization: There is no explicit logic described for filtering or escaping instructions that might be embedded in the source code being analyzed.

openwiki-init