The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to automate the discovery and linting of HTML files within a repository. It employs standard utilities such as find, grep, and xargs to pipe file lists to the htmlhint linter. Evidence: Found in SKILL.md and references/tooling/toolchain.md as the primary mechanism for executing the code gate.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx --no-install to run htmlhint. While npx is a standard Node.js tool, it may fetch the package from the official npm registry if it is not already cached locally. This behavior is consistent with developer workflows for ensuring tool availability without global installation. Evidence: Referenced in SKILL.md and references/tooling/toolchain.md using the npx --no-install htmlhint pattern.
[INDIRECT_PROMPT_INJECTION]: The skill processes local HTML files as part of its linting workflow. While this introduces a surface for indirect prompt injection if an attacker can place malicious instructions within the HTML files, the tools used (htmlhint and prettier) are static analysis utilities and do not execute the content of the files, significantly limiting the risk. 1. Ingestion points: Local .html and .htm files (specified in SKILL.md). 2. Boundary markers: The skill instructs the agent to use authorized paths and specific find/grep patterns to isolate files. 3. Capability inventory: Shell command execution via bash and file write capability via prettier --write. 4. Sanitization: The skill relies on the standard parsing behavior of htmlhint and prettier; no additional sanitization is applied to the raw HTML content before processing.

codeguard-html