agentleFS
skforecastskforecast/skforecast-ai/.claude/skills/verify/SKILL.md

verify

Skill52 starsChanged 47 days agoApache-2.0View on GitHub

Definition of done for skforecast-ai. Lints, runs the tests affected by the current diff and then the full suite, and builds the docs or checks the LLM goldens when those areas changed. Use before reporting any code, test or docs change as finished.

  • Installs packages

189 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

Used with

Projects with it also use instinct-export, instinct-import and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • TheKREATORSSkill

    instinct-export

    Export instincts from project/global scope to a file

    Also in 3 projects that use this

  • TheKREATORSSkill

    instinct-import

    Import instincts from file or URL into project/global scope

    Also in 3 projects that use this

  • AviadPSkill

    risk-assess

    Assess the risk and blast radius of a proposed code change before implementing it. Evaluates CI impact, framework-wide vs localized scope, rollback path, and identifies blindspots. Use when user says "what's the risk", "is this safe", "assess the impact", "risk assess", "blast radius", or before applying framework-wide or CI-affecting changes. Do NOT use for code quality review (use code-quality) or plan validation (use design-validator).

    Also in 3 projects that use this

  • diego-tobalinaSkill

    build-fix

    Fix build and compilation errors systematically. Run when build fails.

    Also in 6 projects that use this

  • AaravChadhaSkill

    secure

    Security review with a confidence gate - a finding exists only with a concrete exploit scenario and a high/medium/low confidence rating; everything else goes to a worth-hardening list, never inflated. Sweeps five surfaces: auth gates, secrets hygiene, injection and unsafe sinks, unsafe deserialization/crypto/transport, and LLM tool-use trust boundaries. Reports only, never fixes. Use when the user asks for a security review or to check vulnerabilities, secrets, or auth.

    Also in 3 projects that use this

  • ArvorCoSkill

    specify

    Create feature specification from natural language description. Use when starting a new feature or creating a spec. Triggers on: create spec, specify feature, new feature spec.

    Also in 10 projects that use this

  • +1 more from TheKREATORS

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 167 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run every step, in order, and report each result.

    line 8
  2. Very specific:

    If a later verify in this session gets the same stamp, report the previous result again instead of rerunning it.

    line 21
  3. Distinctive:

    If no changed file is under one of these paths, nothing can break: report "nothing to verify" with the list of files and stop here.

    line 60
  4. Distinctive:

    Only for quick feedback before the full suite, which runs them again.

    line 113
  5. Very specific:

    Use the worker count that CLAUDE.local.md names instead of auto when it names one.

    line 138
  6. Very specific:

    If goldens differ on purpose, follow /llm-context-change.

    line 148

“Common” is measured: kept by at least 6 in 10 of the 167 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Verify the current change

Run every step, in order, and report each result. Stop at the first failure, fix it, and start again from that step. Never report the task as done with a failing step; if a failure is unrelated to the change, say so explicitly with the output.

Reuse: when a run ends with every step green, note the stamp of the tree, the hash of every file that is not ignored, taken on a copy of the index so the real one stays untouched:

t=$(mktemp); cp "$(git rev-parse --git-path index)" "$t"; GIT_INDEX_FILE="$t" git add -A && GIT_INDEX_FILE="$t" git write-tree; rm -f "$t"

If a later verify in this session gets the same stamp, report the previous result again instead of rerunning it. The stamp does not depend on HEAD, so committing the verified files keeps it (new files too), and any edit changes it.

0. How to run Python

  • Cloud session (CLAUDE_CODE_REMOTE=true): python -m pytest, ruff, mkdocs directly. The SessionStart hook installs the docs extra and exports SKFORECAST_AI_DOCS_PRIVACY=false, so mkdocs build skips the privacy plugin, whose downloads the network policy may block. If mkdocs is missing, install the extra first: pip install -e ".[docs]".
  • Local session: when .claude/settings.local.json sets SKFORECAST_AI_PYTHON_ENV, the SessionStart hook puts that environment on PATH and exports PYTHONPATH=., so python, ruff and mkdocs run directly from the repository root. Otherwise use what CLAUDE.local.md says, and if it does not exist, ask which conda environment to use.

Below, PY stands for that Python command.

1. Scope

git status --short
git diff --name-only HEAD
git ls-files --others --exclude-standard

Also include the commits not yet in the remote release branch (X.Y.x, from __version__ in skforecast_ai/__init__.py), which the commands above no longer show once the work is committed. The same command covers a working branch and unpushed commits on the release branch itself: git diff --name-only origin/<release-branch>...HEAD.

Classify the changed files: package code (skforecast_ai/), tests, docs (docs/, mkdocs.yml, tools/docs/), LLM layer (skforecast_ai/llm/, recommendation/explanation.py, execution/backtesting_runner.py, execution/comparison.py), rendering (skforecast_ai/rendering/).

If no changed file is under one of these paths, nothing can break: report "nothing to verify" with the list of files and stop here.

  • skforecast_ai/, tests/, pyproject.toml
  • docs/, mkdocs.yml, tools/docs/
  • README.md, AGENTS.md, CLAUDE.md, plugin/, .claude-plugin/, packages/, server.json (tests read them)
  • .claude/hooks/

The list names what can break rather than what cannot, so a new folder runs the checks until it is judged safe. Notes and plans in dev/, the other skills and rules in .claude/, and the scripts in tools/ai/, tools/mcp/ and tools/perf/ are outside it: no test or build reads them.

Steps 2 to 4 (lint, affected tests, full suite) run only when a changed file is under skforecast_ai/, tests/ or pyproject.toml. Otherwise skip them and run, for each group with a changed file, only its checks; a change that touches several groups runs the checks of each one:

  • Docs (docs/, mkdocs.yml, tools/docs/): the tests that read the docs, then the docs checks of step 5 (and the release notes check if docs/releases/releases.md changed).

    PY -m pytest tests/test_docs_cli_page.py tests/test_docs_skills_page.py tests/test_mcp_registry_distribution.py tests/tests_mcp/test_skill_md.py -q -p no:cacheprovider
  • Agent instructions and the README (AGENTS.md, CLAUDE.md, README.md):

    PY -m pytest tests/test_source_conventions.py tests/test_mcp_registry_distribution.py -q -p no:cacheprovider
  • Distribution (plugin/, .claude-plugin/, packages/, server.json):

    PY -m pytest tests/test_plugin_distribution.py tests/test_mcp_registry_distribution.py tests/tests_mcp/test_skill_md.py -q -p no:cacheprovider
  • Hooks (.claude/hooks/): the harness check of step 5.

A path added to the list above without a group here runs every step until it gets one.

2. Lint

ruff check skforecast_ai tests

3. Affected tests first

Only for quick feedback before the full suite, which runs them again. Skip this step when this session already ran the affected tests green and no file changed since; running them while iterating is the usual case.

Map each changed module to its tests and run them with -q:

  • skforecast_ai/<sub>/<module>.py -> tests/tests_<sub>/ (for example recommendation/ -> tests/tests_recommendation/).
  • skforecast_ai/assistant.py -> tests/test_assistant_<method>.py for the methods touched.
  • skforecast_ai/cli.py -> tests/test_cli*.py.
  • skforecast_ai/rendering/ -> tests/tests_rendering/ and tests/test_integration_standalone_script.py.
  • Changed test files -> those files.
PY -m pytest <paths> -q -p no:cacheprovider

4. Full suite

PY -m pytest -n auto -q -p no:cacheprovider

Use the worker count that CLAUDE.local.md names instead of auto when it names one. When the docs build of step 5 is due, start it in the background before this step so both run at once.

It must end with no failures and no errors. Warnings are errors in this project, so a new warning fails here too.

5. Conditional checks

  • LLM layer changed: PY -m pytest tests/tests_llm/test_golden_context.py -q. If goldens differ on purpose, follow /llm-context-change.
  • Docs changed: mkdocs build -q -d <scratchpad>/site (never into the repo). Report warnings about broken links or references. Also run PY tools/docs/hooks/page_meta.py: a new or renamed page needs its search engine title and description in tools/docs/hooks/page_meta.yml (CI runs the same check on pull requests).
  • Harness changed (.claude/hooks/): PY -m pytest .claude/hooks -q -p no:cacheprovider (outside testpaths, so the full suite does not run it).
  • User-visible change (API, CLI output, generated scripts, warnings) and no entry in docs/releases/releases.md: add one with /release-note.
  • Release notes changed: PY tools/docs/check_release_notes.py (undefined references, entries over the length ceiling or without a link).

6. Report

A short list: each step, the command, pass or fail, and the test count from the full suite. Mention anything skipped and why.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.