The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The troubleshooting section instructs the agent to install the llm-tldr package from external registries using uv or pip. This introduces a dependency on third-party code that is not pinned or verified within the skill's own manifest.
[DYNAMIC_EXECUTION]: The skill dynamically constructs shell commands by interpolating user-supplied variables such as FOCUS, ENTRY, and PATH (e.g., tldr search "${FOCUS}"). This pattern is susceptible to command injection if the input strings contain shell metacharacters and are not properly sanitized by the execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted codebase content, creating a significant attack surface for indirect prompt injection. Malicious instructions embedded in the code being explored could potentially influence the agent's behavior.
Ingestion points: Codebase files and directory structures accessed via Read, Grep, Glob, and various tldr utility commands.
Boundary markers: The skill does not define specific delimiters or warnings to treat ingested code content as untrusted data.
Capability inventory: The agent has access to powerful tools including Bash, Write, and Task (for spawning sub-agents), which increases the potential impact of a successful injection.
Sanitization: There is no explicit logic for sanitizing or escaping content retrieved from the codebase before it is processed or summarized by the agent.

explore