The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's installer fetches software from several external sources. It clones the OpenSpace framework from the HKUDS (University of Hong Kong) GitHub repository and pins it to a specific commit hash for integrity. It also installs the 'specify-cli' tool from GitHub's official organization repository and various Python packages ('ouroboros-ai', 'cli-anything-hub') from the standard PyPI registry.
[COMMAND_EXECUTION]: The scripts/install.sh and scripts/git-interview-context.sh scripts execute multiple system commands to manage the environment and extract repository metadata. This includes the use of git to analyze logs and churn, pip for package management, and npx to register the skill. The execution is scoped to the installation and primary workflow tasks.
[INDIRECT_PROMPT_INJECTION]: The skill uses repository metadata to inform its Socratic interview process, which creates a potential surface for indirect prompt injection if an attacker can influence the repository's git history.
Ingestion points: The scripts/git-interview-context.sh script aggregates live git logs, commit messages, and file lists into a local file (.ouroboros/interview-context.md) used as context for the agent.
Boundary markers: The generated context file uses Markdown headers and code blocks to structure the repository data, though it does not implement specific adversarial delimiters for the git-generated content itself.
Capability inventory: The skill possesses Bash, Read, and Write capabilities, allowing it to execute the development workflow as defined in the seed specification.
Sanitization: The script performs frequency analysis and basic log formatting (e.g., using sed and sort) but does not apply deep sanitization to the text within commit messages or filenames.

ooo