The Agent Skills Directory

[SAFE]: The skill is designed to assist in the creation and validation of documentation. It performs local file system operations using standard utilities to analyze project metrics such as line counts and file types.
[COMMAND_EXECUTION]: The skill includes suggested shell commands and an auditing script (agents-md-lint.sh) for analyzing local files. These operations are restricted to the project workspace and used for diagnostic purposes like pattern matching and token estimation.
[DATA_EXPOSURE]: While the skill accesses project metadata files (e.g., README.md, package.json, git log), it does not target sensitive directories like .ssh, .aws, or .env files. There are no network-based exfiltration patterns detected.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it reads untrusted content from project files to generate reports.
Ingestion points: Files such as AGENTS.md and README.md are ingested in skills/1-analyze-project/SKILL.md and skills/5-quality-check/SKILL.md.
Boundary markers: The instructions do not define strict boundary markers to prevent the agent from following instructions potentially hidden in the analyzed files.
Capability inventory: Capabilities are limited to file reads and shell-based text analysis (grep, awk, wc).
Sanitization: Content is processed via regular expressions and standard shell pipes without explicit sanitization or escaping.

agents-writer