The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interpolates user-provided search terms directly into shell commands (e.g., npx skills find [query] and clawhub search "[query]") without escaping or validation. This allows a user to execute arbitrary commands by crafting a search query containing shell metacharacters. It also suggests using the -g flag for global installations, which may involve elevated privileges.
[REMOTE_CODE_EXECUTION]: The skill's primary purpose is to install and execute code from third-party registries such as skills.sh, clawhub.com, and tessl.io. It specifically encourages installation methods like npx skills add <owner/repo@skill> and tessl install github:user/repo which execute code from arbitrary GitHub repositories.
[EXTERNAL_DOWNLOADS]: The skill uses curl to fetch content from tessl.io and npm to install global CLI tools. While functional, this involves downloading and executing content from external servers at runtime.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes and displays output from external registries that may contain malicious instructions.
Ingestion points: Output from searches on skills.sh, clawhub, and tessl.io ingested into the agent context via SKILL.md.
Boundary markers: None; results from external sources are presented without delimiters or instructions to ignore embedded prompts.
Capability inventory: Shell command execution, package installation, and network access (curl) as defined in SKILL.md.
Sanitization: There is no evidence of validation or filtering for content retrieved from external registries.

unified-find-skills