The Agent Skills Directory

[SAFE]: The skill implements a routing pattern to direct user requests to specific AlterLab academic tools using a localized skill index. It does not access external repositories or unverified code.- [SAFE]: Multi-stage workflows (orchestration mode) require mandatory user clarification and confirmation of the plan and cost before execution, providing a strong human-in-the-loop safeguard against unintended tool execution.- [SAFE]: All referenced dependencies and skill names are consistently within the author's vendor namespace ('alterlab-*') as defined in the skill index.- [SAFE]: The skill uses a restricted set of tools ('Read Task') and does not require network access, external API keys, or privileged system access.- [PROMPT_INJECTION]: The skill processes untrusted user input to select and orchestrate tools from its index (Category 8 surface). 1. Ingestion points: User requests in the main prompt. 2. Boundary markers: No explicit delimiters are used for user input in the instructions. 3. Capability inventory: Dispatches tasks to other skills which may have file-system or command execution capabilities (e.g., 'alterlab-blast'). 4. Sanitization: Mitigation is handled by a mandatory 'Clarify' and 'Confirm' step requiring the user to approve the orchestration plan before execution. The risk is assessed as safe given the architectural constraints and human review requirement.

alterlab-skill-finder