The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to run markdown diagnostic tools.
Evidence: It executes npx --no-install markdownlint-cli2 --version and npx --no-install markdownlint-cli2 as part of its core workflow defined in the Execution Contract and SKILL.md workflow steps.
Context: The use of --no-install is a security-positive pattern that ensures the agent only uses locally available tools and does not fetch arbitrary packages from the network during execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from Markdown files which could be used to host adversarial instructions aimed at the agent.
Ingestion points: Reads files with .md and .markdown extensions within the target repository during the linting process.
Boundary markers: The skill does not implement explicit boundary delimiters or warnings when reading file content for the agent to process.
Capability inventory: The skill has the capability to execute shell commands (npx), read file content, and write modifications back to the file system using the --fix flag.
Sanitization: The risk is partially mitigated by the Rules section which mandates that all automatic fixes must have their diffs reviewed by the user and that the agent should not perform "batch guessing" of semantic fixes.

codeguard-markdown