The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of local shell commands via a specialized CLI to perform professional searches, authentication checks, and data retrieval.
Evidence: Multiple instructions in SKILL.md and references/cli-reference.md guide the agent to use commands such as lessie find-people, lessie enrich-org, and lessie status.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to check for updates and installs external packages from well-known registries to establish its execution environment.
Evidence: SKILL.md includes commands to fetch version metadata from GitHub (raw.githubusercontent.com/LessieAI/lessie-skill) and installs the @lessie/cli package from the NPM registry.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it ingests and processes data from external web sources and social profiles.
Ingestion points: Data retrieved via web_search, web_fetch, and review_people tools from the open web and social media platforms as described in SKILL.md and references/workflow-patterns.md.
Boundary markers: No explicit boundary markers or instructions to disregard embedded commands are defined for the processing of external content.
Capability inventory: The agent can execute shell commands via the lessie CLI, perform network operations, and write configuration data to the local file system at ~/.lessie (documented in SKILL.md).
Sanitization: No specific sanitization, filtering, or validation logic is detailed for the external content before it is processed by the agent.
[DYNAMIC_EXECUTION]: The skill includes a fallback mechanism that allow the agent to dynamically discover and invoke remote tools by querying their schema at runtime.
Evidence: references/cli-reference.md instructs the agent to use lessie tools to fetch the current remote schema and lessie call to execute tools with raw JSON arguments if standard command shortcuts fail to execute correctly.

people-search