The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts and standard system utilities (such as jq, git, and grep) to diagnose the .woostack/ workspace. These operations are restricted to local file reads and do not involve network connectivity or credential access.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from the repository's .woostack/config.json and .woostack/config.local.json files, creating a surface for indirect prompt injection.
Ingestion points: Configuration data is read and parsed by scripts/config/resolve-config.sh.
Boundary markers: The SKILL.md instructions establish a 'HARD GATE,' mandating explicit user approval before any workspace repairs are applied, and emphasize that the diagnostic engine is provider-free.
Capability inventory: The skill's primary capability is proposing and applying local file mutations to the .woostack/ directory. It explicitly avoids remote operations, scaffolding, or credential management.
Sanitization: The resolve-config.sh script performs automated validation using jq, which includes checking for credential-like keys via regex and enforcing a specific JSON schema for GitHub-related policies.

woostack-doctor