The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user-provided HTML and Markdown UI specifications, which serves as a potential vector for indirect prompt injection.
Ingestion points: Artifact content is processed by scripts/checks.py and manually reviewed by the agent as part of the four-pass protocol described in SKILL.md.
Boundary markers: SKILL.md establishes a structured four-pass protocol (Inventory, Criterion sweep, Severity assignment, Envelope assembly) intended to maintain analytical rigor.
Capability inventory: The skill's scripts (checks.py and merge.py) perform file reading and local JSON processing.
Sanitization: There is no documented evidence of input sanitization or character escaping for the analyzed artifact text.
[EXTERNAL_DOWNLOADS]: The skill protocol includes the installation of a common external dependency.
Evidence: SKILL.md instructs the user to run pip install "jsonschema>=4.20,<5" to support the automated validation logic. jsonschema is a well-known and reputable library.
[DYNAMIC_EXECUTION]: The test suite employs dynamic module loading to facilitate unit testing of local scripts.
Evidence: scripts/tests/test_checks_usability.py uses importlib.util to load the checks.py module. This is a standard and acceptable practice for testing scripts located in non-package directories.

critique-usability