Skip to content

Adonis0123/adonis-skills / .agents/skills/repo-skill-creator

repo-skill-creator

Verified

Create and maintain repository skills for adonis-skills. Use when users ask to scaffold a new skill under skills/, generate agents/openai.yaml, run skill validation, refresh web index data, or standardize the skill authoring workflow.

evals/evals.json

View on GitHub
{
  "skill_name": "repo-skill-creator",
  "evals": [
    {
      "id": 1,
      "prompt": "Use repo-skill-creator to validate the existing skills in this repository. There is no new skill to create. Do not stage anything.",
      "expected_output": "Runs repository skill validation without calling skills:new, skills:finalize:new, or git add.",
      "files": [],
      "expectations": [
        "Classifies the request as validation rather than creation.",
        "Does not create a skill or modify the Git index."
      ]
    },
    {
      "id": 2,
      "prompt": "Finalize skills/example-skill, but leave all files unstaged.",
      "expected_output": "Runs the path-finalize pipeline for the named skill and does not stage files.",
      "files": [],
      "expectations": [
        "Uses the explicitly named skill path.",
        "Does not run git add."
      ]
    },
    {
      "id": 3,
      "prompt": "Prepare all newly added skills for commit: finalize them and stage only their files plus the generated skill indexes.",
      "expected_output": "Uses finalize-and-stage mode and recognizes both current generated index files.",
      "files": [],
      "expectations": [
        "Stages only discovered skill directories and changed generated index files.",
        "Uses skills-index-lite.json and skills-detail-index.json, not the removed skills-index.json."
      ]
    },
    {
      "id": 4,
      "prompt": "In the provided isolated repository, create skills/csv-expense-totals with author Platform #1. Implement a Python standard-library CSV helper that groups by currency and sums amounts with Decimal, including 0.1 plus 0.2. Reject missing columns, blank, invalid and non-finite amounts without partial stdout or echoing private input values. Add meaningful tests and a usable example, then run the actual finalize pipeline and read back the generated skill, helper, usage and both indexes. Leave everything unstaged and preserve unrelated inputs.",
      "expected_output": "Creates a usable skill, preserves the quoted author, passes behavior tests and the complete finalize pipeline, and verifies actual indexed output.",
      "files": [],
      "expectations": [
        "Executes real initialization, implementation tests and quick-validate, validate and index in order.",
        "Passes the same independent precision, grouping, missing-column and invalid-amount cases as the baseline, including invalid data after a valid row with no partial output.",
        "Preserves unrelated skill and tool inputs and verifies author metadata and both actual index schemas.",
        "Uses implementation reads when needed rather than assuming every normal creation needs every tool implementation.",
        "Compare native elapsed time and terminal usage under the same model and configuration; fewer source characters alone are not an efficiency result."
      ]
    },
    {
      "id": 5,
      "prompt": "In the provided isolated repository, finalize the existing skills/csv-expense-totals. Do not edit any existing skill or tool, stage files, install dependencies or change permissions. Use the real repository pipeline and report its actual outcome; diagnose a failure from the available evidence without repairing unrelated content.",
      "expected_output": "With a valid target and an invalid unrelated existing skill in the fixture, target quick-validation succeeds, repository validation fails, and index generation does not run. The report names the real validation problem without unauthorized repair.",
      "files": [],
      "expectations": [
        "Runs the actual target finalize command once and preserves its nonzero validation outcome.",
        "Reads relevant source or metadata when the failure needs diagnosis; the normal-path loading rule must not prevent that investigation.",
        "Does not run index generation after failed repository validation, modify unrelated skills, stage files or claim success.",
        "All skill and tool inputs remain unchanged; only authorized report and task-local validation cache outputs may change."
      ]
    }
  ]
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall