The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones the open-gsd/gsd-core repository from GitHub into a temporary directory for auditing. This repository is an external source not associated with the skill author's known domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the external repository to generate new capabilities or update existing ones.
Ingestion points: Cloned upstream files (GSD_ROOT), specifically changeset markdown files and capability source code.
Boundary markers: Absent; the instructions do not specify delimiters or warnings to isolate external content.
Capability inventory: Execution of shell scripts, use of GitHub CLI for issue management, and file system write access.
Sanitization: Absent; the skill relies entirely on the agent's reasoning and manual maintainer review to filter malicious content.
[PRIVILEGE_ESCALATION]: The skill executes PowerShell scripts using the -ExecutionPolicy Bypass flag, which is a technique used to override security restrictions for script execution on Windows.
[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) to read and update repository issue bodies for persistence and performs multiple shell operations to compute deltas between commits.
[DYNAMIC_EXECUTION]: The skill constructs file paths and shell commands dynamically based on output from previous steps, such as the GSD_ROOT variable and SHA identifiers retrieved from git history.

sync-gsd