The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses aggressive behavioral overrides and persona reinforcement techniques, such as the "OWNER OPERATING DIRECTIVE" with "ABSOLUTE" importance and a policy that "supersedes-all-other-sections." These patterns are designed to constrain the agent's reasoning and prioritize specific operational rules over general system guidelines.
[COMMAND_EXECUTION]: The skill automates repository management and package publishing through the execution of shell commands like git push, npm version patch, and npm publish. While these are expected for a DevOps skill, they represent significant capabilities that require careful oversight.
[EXTERNAL_DOWNLOADS]: The skill references external resources like the mksglu/context-mode repository on GitHub and suggests installation via npx skills add. These references are to well-known services and the vendor's own infrastructure.
[PROMPT_INJECTION]: The skill processes untrusted input from GitHub issues and pull requests, creating a surface for Indirect Prompt Injection (Category 8).
Ingestion points: External content from GitHub issues and PRs fetched via gh CLI commands in triage-issue.md and review-pr.md.
Boundary markers: The skill lacks explicit delimiters or specific instructions to ignore embedded commands when processing external text.
Capability inventory: The skill can spawn multiple subagents with "ultrathink" authority, merge PRs, and publish packages to npm.
Sanitization: The skill relies on manual "Claim Verification" gates and "Architect" review processes to validate external behavioral claims before implementation.

context-mode-ops