The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs a comprehensive suite of standard Unix utilities including ls, grep, find, tree, cat, and wc to perform reconnaissance on the target codebase. These operations are used as intended for identifying entry points, mapping directory structures, and analyzing dependency manifests like package.json, go.mod, and pyproject.toml.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted data (the contents of the codebase being analyzed). While this presents a theoretical surface for indirect prompt injection if malicious instructions are embedded in the analyzed files, the skill relies on deterministic shell commands for its analysis phase, which significantly limits the risk of the agent being manipulated by the data it reads.
Ingestion points: Reads local files and directory structures via shell commands in exploration-protocol.md and tech-stack-detection.md.
Boundary markers: None explicitly defined in the shell scripts, though the response-format.md provides a structured template for the final report.
Capability inventory: Limited to file system read operations and shell-based text processing across all reference scripts.
Sanitization: Standard shell utilities are used; no explicit escaping of file content is performed before presentation in the report.

exploration