The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses Bun's shell execution wrapper ($) and bash scripts to automate development tasks. This includes git repository initialization, commit and push operations, file system management (creating directories, symlinks, and copying files), and GitHub CLI commands for repository creation and management.
[EXTERNAL_DOWNLOADS]: The primary function of the skill is to fetch external repositories from GitHub using the ghq tool. These downloads are directed from a well-known service (github.com) and are central to the skill's operation for studying and developing on external codebases.
[DATA_EXFILTRATION]: The skill accesses the GitHub API to retrieve the current user's login name and organization list. This information is used for indexing and scoring repositories within the user's workspace but represents access to account-level metadata.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it scans and indexes markdown files (.md) from external repositories during its 'reunion' and 'index' tasks. Malicious instructions embedded in these external files could potentially be ingested into the agent's context.
Ingestion points: scripts/reunion.ts (scans cloned directories like docs/ and learnings/) and scripts/index.ts (processes generated manifests).
Boundary markers: No explicit delimiters are used when processing the content of these files.
Capability inventory: The skill has significant local and network capabilities, including shell access and GitHub API interaction.
Sanitization: External file content is read and logged without specialized sanitization for AI instructions.

project