The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and execute an installation script from the official domain https://claude.ai/install.sh. While the curl | sh pattern is high-risk in unknown contexts, the source is a recognized well-known service provider.
[PROMPT_INJECTION]: The skill defines workflows that ingest untrusted external data from Git Pull Requests and source code repositories. This establishes an attack surface for indirect prompt injection if the ingested content contains malicious instructions.
Ingestion points: Git PR data via gh pr checkout and code analysis using tools like Grep or @src/ references.
Boundary markers: None are specified to delimit external content from instructions.
Capability inventory: The skill has access to powerful tools including Bash, Write, and project-wide file access.
Sanitization: No sanitization or validation logic is proposed for handling external data before it is processed by the agent.
[COMMAND_EXECUTION]: The instructions document the usage of shell execution capabilities and immediate command execution syntaxes (e.g., !cmd). While these are standard features of the productivity tools being taught, they provide the agent with broad execution power over the local environment.

agentic-workflow