huggingface/inspect_evals
Collection of evals for Inspect AI
https://skillcdn.ai/gh/huggingface/inspect_evalsConnect this address to your AI to use these skills. How to connect
- Unverified
- Default branch
main - Commit
c876ff7 - License
MIT
Explore this repository
Searches the original text of skills and documents; display translations are not searched.
Skills
- build-repo-contextCrawl repository PRs, issues, and review comments to distill institutional knowledge into a shared knowledge base. Run periodically by "context agents" to maintain agent_artefacts/repo_context/REPO_CONTEXT.md. Trigger only on specific request.MIT
- check-trajectories-workflowUse Inspect Scout to analyze agent trajectories from evaluation log files. Runs default and custom scanners to detect external failures, formatting issues, reward hacking, and ethical refusals. Use when user asks to check/analyze agent trajectories. Trigger when the user asks you to run the "Check Agent Trajectories" workflow.MIT
- ci-maintenance-workflowCI and GitHub Actions maintenance workflows — fix a failing test from a CI URL, add @pytest.mark.slow markers to slow tests, or review a PR against agent-checkable standards. Use when user asks to fix a failing test, mark slow tests, or review a PR. Trigger when the user asks you to run the "Write a PR For A Failing Test", "Mark Slow Tests", or "Review PR According to Agent-Checkable Standards" workflow.MIT
- code-quality-fix-allFix code quality issues identified in a code quality review stored in agent_artefacts/code_quality/<topic>/. Systematically addresses issues found by the code-quality-review-all skill for ANY code quality topic, with validation and testing at each step. Use when user asks to fix issues from a code quality review, or asks to fix issues from agent_artefacts/code_quality/<topic>.MIT
- code-quality-review-allReview all evaluations in the repository against a single code quality standard. Checks ALL evals against ONE standard for periodic quality reviews. Use when user asks to review/audit/check all evaluations for a specific topic or standard. Do NOT use for reviewing a single eval (use eval-quality-workflow instead) or for test coverage (use ensure-test-coverage instead).MIT
- create-evalImplement a new evaluation from an issue or paper. Guides through requirements gathering, architecture design, scaffolding, implementation, testing, and quality checks. Use when user asks to create/implement/build a new evaluation.MIT
- ensure-test-coverageEnsure test coverage for a single evaluation - both reviewing existing tests and creating missing ones. Analyzes testable components, checks tests against repository conventions, reports coverage gaps, and creates or improves tests. Use when user asks to check/review/create/add/ensure tests for an eval. Use whenever you are asked to review an evaluation that contains tests, or whenever you need to write a suite of tests. Do NOT use for fixing a specific failing CI test (use ci-maintenance-workflow instead).MIT
- eval-quality-workflowFix or review a single evaluation against all EVALUATION_CHECKLIST.md standards. Use "fix" mode to refactor an eval into compliance, or "review" mode to assess compliance without making changes. Use when user asks to fix, review, or check an evaluation's quality. Trigger when the user asks you to run the "Fix An Evaluation" or "Review An Evaluation" workflow. Do NOT use for reviewing ALL evals against a single code quality standard (use code-quality-review-all instead).MIT
- eval-report-workflowCreate an evaluation report for a README by selecting models, estimating costs, running evaluations, and formatting results tables. Use when user asks to make/create/generate an evaluation report. Trigger when the user asks you to run the "Make An Evaluation Report" workflow.MIT
- investigate-datasetInvestigate datasets from HuggingFace, CSV, or JSON files to understand their structure, fields, and data quality. Trigger whenever you need to explore or inspect a dataset yourself without using pre-written scripts.MIT
- prepare-releasePrepare a new release of inspect_evals by creating a release branch, collecting changelog fragments, and opening a PR. Use when user asks to cut/prepare/create a new release or version bump.MIT
- prepare-submission-workflowPrepare an evaluation for PR submission — add dependencies, tests, linting, eval.yaml, README, and changelog. Use when user asks to prepare an eval for submission or finalize a PR. Trigger when the user asks you to run the "Prepare Evaluation For Submission" workflow.MIT
- read-eval-logsView and analyse Inspect evaluation log files using the Python API. Trigger whenever you need to look at a .eval file yourself without using pre-written scripts.MIT
- write-an-adrWrite an Architectural Decision Record (ADR) to document a significant design choice. Use when user asks to write/create/document an ADR, or to record an architectural decision.MIT