The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands including git, ls, and cat to retrieve local metadata, commit history, and version information from the project's directory structure.
[COMMAND_EXECUTION]: It attempts to run a local TypeScript file located at src/skills/oracle-family-scan/scripts/fleet-scan.ts using the bun runtime to perform a fleet scan.
[EXTERNAL_DOWNLOADS]: Fetches remote data from GitHub using the gh CLI tool, specifically retrieving repository lists and issue bodies from the Soul-Brews-Studio organization to display live project statistics.
[PROMPT_INJECTION]: The skill processes untrusted content from a public GitHub issue (issue #60 in the oracle-v2 repository). This creates an indirect prompt injection vulnerability where an attacker could influence the agent's narrative by modifying the external issue content.
Ingestion points: Data is ingested from the output of gh issue view 60 --repo Soul-Brews-Studio/oracle-v2 within SKILL.md.
Boundary markers: There are no explicit markers or safety instructions used to distinguish the fetched issue data from the agent's primary instructions.
Capability inventory: The skill possesses the ability to execute shell commands, read local files, and perform network requests via the GitHub CLI across its defined logic.
Sanitization: While the output is piped through basic filters like grep and head, there is no sanitization to prevent the AI from interpreting instructions embedded in the issue text.

about-oracle