The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the Bun installation script from https://bun.sh/install. This reference is considered safe as it targets a well-known service and incorporates SHA-256 integrity verification to ensure the script has not been tampered with.
[DYNAMIC_EXECUTION]: The skill uses eval to execute output from gstack-slug and employs process substitution to source output from gstack-repo-mode. These binaries are located within the user's local environment at ~/.claude/skills/gstack/bin/. While these appear to be part of the skill's own tooling framework, dynamic execution of locally computed logic is a significant capability.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external, potentially untrusted web applications using a browser tool to generate reports. Malicious instructions embedded in a web page's HTML, metadata, or console logs could influence the agent's behavior during the testing process.
Ingestion points: Browser tool outputs including page snapshots, link maps, and console error logs.
Boundary markers: None present in the instructions to separate untrusted web content from the agent's instructions.
Capability inventory: The skill possesses shell access (Bash), file system write access (Write), and network access (WebSearch).
Sanitization: No specific sanitization or filtering of external data is described before the agent processes it for reporting.
[COMMAND_EXECUTION]: Executes various Git commands (git diff, git log, git commit) to analyze the project's state and automatically update CLAUDE.md with skill routing rules based on user preferences.

qa-only