The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file contains a 'Version Self-Check' section that instructs the AI agent to autonomously execute shell commands (git rev-parse HEAD and git ls-remote origin HEAD) to compare local and remote versions. It also directs the agent to perform file system writes to a .last-update-check file.
[REMOTE_CODE_EXECUTION]: The skill encourages remote code execution through its installation and update procedures. The README.md suggests using npx skills add alchaincyf/steve-jobs-skill, which executes remote packages. The SKILL.md update logic provides instructions for a git pull operation, which downloads and executes remote skill content.
[DATA_EXFILTRATION]: The automated git ls-remote instruction transmits repository metadata (branch and commit hashes) from the local environment to a remote server (GitHub) during the version check process.
[INDIRECT_PROMPT_INJECTION]: The skill defines a mandatory research workflow in SKILL.md (Step 2) that ingests untrusted data from the web to analyze products and user feedback.
Ingestion points: The WebSearch tool output is interpolated into the agent context during the '乔布斯式研究' (Jobs-style research) phase.
Boundary markers: Absent. There are no specific delimiters or instructions to ignore embedded commands within the fetched web content.
Capability inventory: The agent possesses shell execution and file system write capabilities as demonstrated by the Git-based update and version tracking logic.
Sanitization: Absent. The skill does not provide mechanisms to filter or escape the content retrieved from external websites before processing it within the 'Jobs' persona.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and update information from the vendor's repository on GitHub. These references to github.com/alchaincyf/* and vercel-labs/skills represent the intended management of the skill's lifecycle.

steve-jobs-perspective