The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided projects and artifact manifests to run functional tests. If an attacker provides a project containing malicious test code, the skill will execute it with the agent's permissions.
Ingestion points: The skill reads project configurations and source files defined in .styleseed/artifacts/ and .styleseed/manifests/ (e.g., in scripts/run-functional-tests.mjs).
Boundary markers: While the skill checks that test files are within implementation source roots, it does not sanitize or validate the logic within the tests themselves.
Capability inventory: The skill uses child_process.spawnSync in scripts/run-functional-tests.mjs to execute Node.js tests and in scripts/evidence-gate.mjs to run Git commands.
Sanitization: No sanitization is performed on the Javascript content executed during functional testing.
[COMMAND_EXECUTION]: The skill executes external CLI tools (Node.js and Git) to perform its core functions.
Evidence: scripts/run-functional-tests.mjs spawns process.execPath (Node) to run project tests. scripts/evidence-gate.mjs spawns git to verify repository revisions.
[DYNAMIC_EXECUTION]: The skill dynamically executes Javascript test files located within the project's source tree.
Evidence: The runFunctionalTests function in scripts/run-functional-tests.mjs identifies and executes .mjs files from the project's source roots using the Node.js test runner.

ss-score