The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends an installation method that pipes a remote shell script directly into bash. This pattern allows for arbitrary code execution on the user's system without prior verification of the script's content.
Evidence: curl -fsSL https://raw.githubusercontent.com/QuinsZouls/code-index/master/install.sh | bash in references/management.md.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install external software from GitHub repositories and releases if the tool is missing.
Evidence: Mentions of go install github.com/QuinsZouls/code-index/src@latest and manual downloads of binary assets from GitHub Releases in references/management.md.
[COMMAND_EXECUTION]: The skill is designed to have the agent autonomously execute shell commands to manage a local database index, including initialization, indexing, and search operations.
Evidence: Instructions in SKILL.md to run codeindex init, codeindex index, and codeindex search automatically without user intervention.
[PROMPT_INJECTION]: The skill indexes and searches local codebase content, which introduces a surface for indirect prompt injection if source files contain malicious instructions.
Ingestion points: codeindex index reads all files matching the project's inclusion patterns (e.g., .py, .js, .go, .md).
Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are provided when search results are returned to the agent.
Capability inventory: The agent has the capability to execute shell commands and read files, which could be abused if the agent follows instructions hidden within indexed code.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the codebase index.

codeindex