The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes a setup routine in the 'SETUP' section that downloads and executes an installation script for the Bun runtime from 'https://bun.sh/install'. This script is used to provision the environment if the required runtime is missing.
[EXTERNAL_DOWNLOADS]: Fetches the official Bun installation script from the well-known 'bun.sh' service. The process includes a SHA-256 checksum verification ('actual_sha=$(shasum -a 256 "$tmpfile" | awk '{print $1}')') to ensure the integrity of the downloaded file.
[COMMAND_EXECUTION]: The preamble and various phases execute multiple local binaries and scripts located in '~/.claude/skills/gstack/bin/', such as 'gstack-update-check', 'gstack-config', 'gstack-slug', and 'gstack-telemetry-log'. These are internal utility components for the gstack tool suite.
[DYNAMIC_EXECUTION]: Uses 'eval "$(~/.claude/skills/gstack/bin/gstack-slug 2>/dev/null)"' in the preamble and context recovery sections to dynamically load environment variables (like 'SLUG') provided by local helper binaries.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted data from external URLs by reading console errors, page links, and text content via the browse daemon ('$B console', '$B links', '$B text') during monitoring phases.
Boundary markers: Not explicitly defined in the prompts; the agent is instructed to compare these inputs against a stored baseline.
Capability inventory: The skill has access to Bash, Read, and Write tools which could be targeted by instructions embedded in the monitored web content.
Sanitization: Browser-ingested content is used for comparison and reporting; however, explicit sanitization logic for terminal output of this content is not detailed.

canary