The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's preamble executes a wide array of local utility binaries (e.g., gstack-config, gstack-slug, gstack-timeline-log) and shell commands to manage configuration, sessions, and telemetry. It uses dynamic execution via eval and source on the output of local vendor tools.
[DATA_EXFILTRATION]: The skill includes logic for synchronizing project artifacts to a remote repository and logging usage telemetry. These features are designed to trigger only after explicit user consent is obtained via the AskUserQuestion tool.
[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests untrusted data from external websites via the $B (browser) toolset. This content is processed by the AI to extract structured information, posing a risk of indirect prompt injection if the source content contains malicious instructions.
Ingestion points: Web page content (text and HTML) accessed via $B snapshot and $B html in SKILL.md.
Boundary markers: No specific delimiters or instructions are provided to separate scraped content from the agent's internal instructions.
Capability inventory: The skill allows access to powerful tools including Bash and Read.
Sanitization: There is no evidence of explicit sanitization or filtering applied to the retrieved web data before processing.

scrape