The Agent Skills Directory

[COMMAND_EXECUTION]: The script .claude/skills/blueprintkit/webapp-testing/scripts/with_server.py utilizes subprocess.Popen with shell=True to execute arbitrary commands for starting and managing development servers. This provides a broad shell execution surface that could be misused.
[COMMAND_EXECUTION]: The script .claude/skills/blueprintkit/web-artifacts-builder/scripts/init-artifact.sh modifies the global system environment by executing npm install -g pnpm if the package manager is not found during project initialization.
[EXTERNAL_DOWNLOADS]: Multiple automation scripts in the web-artifacts-builder and ci-cd-pipeline-builder skills download and install a large number of external dependencies from public registries (npm) during project setup and bundling operations.
[PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection. 1. Ingestion points: The webapp-testing skill reads raw content from web pages using Playwright's page.content(). 2. Boundary markers: Absent; there are no explicit instructions to the agent to treat the ingested web content as untrusted or to ignore instructions embedded within it. 3. Capability inventory: The skill possesses significant command execution capabilities via with_server.py and init-artifact.sh. 4. Sanitization: Absent; ingested data is used directly for element discovery and reconnaissance without validation or escaping.

blueprintkit