The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes a setup routine that downloads the Bun installer from its official domain.
The process uses curl to fetch the installation script and bash to execute it.
Security controls include a version pin (1.3.10), a hardcoded SHA-256 checksum verification (BUN_INSTALL_SHA), and a mandatory user confirmation step before the download begins.
[EXTERNAL_DOWNLOADS]: Fetches installation scripts and assets from bun.sh, which is recognized as a well-known developer service.
[DYNAMIC_EXECUTION]: The preamble uses eval and source on the output of local helper binaries (such as gstack-slug and gstack-repo-mode) to dynamically manage environment state and project-specific configuration.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes data from external, potentially attacker-controlled sources.
Ingestion points: Reads pull request titles, bodies via the gh CLI, and commit history via git log commands.
Boundary markers: Extracted metadata is displayed within ASCII reports; however, the skill does not explicitly sanitize this content to prevent embedded instructions from influencing subsequent agent actions.
Capability inventory: The skill possesses high privileges, including full Bash tool access, the ability to merge/delete/revert PRs, and browser automation capabilities.
Sanitization: No specific filtering or validation of PR metadata is observed before it is added to the agent's context.
[COMMAND_EXECUTION]: Extensively utilizes the Bash tool to perform system-level operations, including repository management, GitHub CLI interactions, and automated browser testing for deployment health checks.
[DATA_EXFILTRATION]: Implements a telemetry system that collects skill usage and duration data.
The process is transparently described to the user and requires an explicit opt-in via the AskUserQuestion tool before any data is logged or transmitted.

land-and-deploy