The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a behavioral shift instructing the agent to 'plow ahead' and bypass routine clarification stops. While this reduces human-in-the-loop oversight, the instructions include explicit safety constraints. The agent is directed to stop for 'true blockers' such as unavailable credentials, destructive or irreversible actions, and high-risk safety, privacy, or security uncertainties.
[COMMAND_EXECUTION]: The instructions empower the agent to autonomously 'run targeted validation' and 'fix issues found by validation' during its work loop. This involves executing shell commands or scripts to verify implementation choices without immediate user supervision.
[EXTERNAL_DOWNLOADS]: The README provides installation instructions using npx @agent-native/skills@latest, which fetches the skill from the platform's official package registry.
[INDIRECT_PROMPT_INJECTION_SURFACE]: The skill increases exposure to indirect prompt injection by encouraging the agent to process external data (files, docs, PRs, issues) autonomously.
Ingestion points: The agent is instructed in SKILL.md to inspect 'real files, docs, issue, PR, screenshots, or runtime behavior' before editing.
Boundary markers: The skill does not define specific delimiters or warnings for the agent to ignore instructions embedded within these external sources.
Capability inventory: The agent has the capability to implement file changes, execute validation commands, and spawn subagents for research or implementation as per SKILL.md.
Sanitization: There are no explicit instructions for the agent to sanitize or escape content retrieved from external sources before processing.

plow-ahead