Files
Bo 3213afcf1c Default to native execution and report independently accepted work (#1129)
## Change

Make native coding-agent execution the default AgentOps entry path with
zero mandatory skills. Preserve full bundles and add repeatable `ao
skills link --skill NAME` selection, validating the entire selection
before writes. Align product, installation, architecture and generated
command documentation.

Extend the existing trial readout to separate endpoint test results,
execution state and independently accepted work. Bind supplied judgments
to exact content, acceptance and native evidence. Reject empty
implementation subjects and require the caller's complete criterion ID
set before reporting acceptance. Preserve genuine nonempty and
deletion-only subjects, valid failures and missing-proof outcomes.

## Validation

- Native onboarding from empty home/consumer directories produces no
setup files; selective/full linking and failure boundaries are covered.
- Actual RED/GREEN regressions cover empty subjects and the
partial-criterion omission found by independent review.
- Full Go build, vet and race/shuffle tests; affected Go lint; 88 Python
readout/statistics tests passed.
- All 73 gates, generated projections, strict documentation build and
local aggregate passed (10 passed; one documented optional absence).
- All nine PR checks succeeded at
`7df0d42b12f35ffc22008cc10a40339afcfbb6a0`.
- Fresh author-distinct review passed all six acceptance criteria over
all 59 changed paths, with no findings or unchecked scope, after
repairing the criterion-coverage finding.

## Evidence limits

The real native coding repair demonstrates usability, not comparative
skill uplift. The strict live-session machine replay remains NOT_PROVEN
where execution/identity observations are unavailable; the source review
PASS is retained separately. Existing cohort limits and the historical
aggregate-enforcement gap remain unwaived. No new comparative cohort,
scheduler, skill-corpus deletion, memory migration or global
installation is included.
2026-09-10 20:28:22 +00:00

6.6 KiB

Contributing to AgentOps

AgentOps is the operational layer for coding agents. Contributions are welcome across docs, runtime behavior, CLI reliability, tests, and skills.

If you want the fastest path to a meaningful first contribution, start here:

Getting Started

Prerequisites

  • GitHub account
  • Git
  • A supported runtime if you want to test installs end to end (Codex, Claude Code, OpenCode, or another skill-install target)
  • Comfort editing Markdown and YAML frontmatter

Fork and Clone

git clone https://github.com/YOUR_USERNAME/agentops.git
cd agentops

High-Leverage Ways To Contribute

You do not need to add a brand-new skill to make a good contribution.

Useful contribution paths:

  • Docs clarity: tighten README, guides, examples, or onboarding
  • Skill quality: improve an existing SKILL.md, references, or validation scripts
  • Runtime reliability: hooks, install paths, and lifecycle behavior
  • CLI ergonomics: help text, JSON output, workflow plumbing
  • Validation and CI: tests, parity checks, and failure-proofing

First Contribution In 30 Minutes

  1. Pick one narrow improvement: a typo, broken link, unclear instruction, or stale command example.
  2. Create a branch: git checkout -b docs/first-contribution-fix
  3. Make the change.
  4. Run the narrowest relevant check.
  5. Open a PR with before/after context.

Adding A Skill

Use Create Your First Skill for the full walk-through.

The short version:

  1. Create skills/your-skill-name/.
  2. Add a current SKILL.md with skill_api_version: 1.
  3. Keep the entry point lean; put deeper material in references/ only when needed.
  4. Link every references/*.md file from SKILL.md.
  5. Run the local gates before opening the PR.

Local Validation

At minimum, run:

# Skill structure and reference integrity
bash skills/skill-builder/scripts/heal.sh --strict

# Docs, links, and skill-count consistency
bash tests/docs/validate-doc-release.sh

If you add or remove a skill directory, you must run:

python3 scripts/generate-skill-mesh.py

This updates the skill count across SKILL-TIERS.md, PRODUCT.md, README.md, docs/SKILLS.md, docs/ARCHITECTURE.md, and using-agentops/SKILL.md. The doc-release-gate CI job fails if counts drift, so skipping this step will block your PR. If you're unsure whether your change affects counts, run the script anyway — it's idempotent when counts are already in sync.

If you touched Codex-facing behavior or checked-in Codex artifacts, also run:

bash scripts/audit-codex-parity.sh --skill your-skill-name
bash scripts/validate-codex-generated-artifacts.sh --scope worktree

For a fast changed-surface check, run:

ao gate check --fast --scope worktree

Working On The Docs Site

The published site at boshu2.github.io/agentops is built with MkDocs Material, not Jekyll. If your change touches anything under docs/, the top-level README.md, CHANGELOG.md, or skills/**/SKILL.md, verify the site still builds:

# Strict build (what CI runs). First run creates .venv-docs/ and installs
# the pinned toolchain from requirements-docs.txt (uv preferred, pip fallback).
scripts/docs-build.sh --check

# Live-reload dev server at http://127.0.0.1:8000
scripts/docs-build.sh --serve

MkDocs-specific expectations:

  • Every internal link must resolve. mkdocs build --strict fails on unresolved relative links; tests/docs/validate-links.sh catches the same class without a Python toolchain. --check tolerates only the enumerated, intentional cross-references in tests/docs/mkdocs-strict-allowlist.txt (docs pages linking to real repo artifacts outside the site) — see the strict-allowlist policy; any other strict warning fails the build, so fix the link rather than silencing it.
  • Skill pages and the CLI reference are generated at build time from skills/**/SKILL.md and cli/docs/COMMANDS.md respectively — do not hand-author docs/skills/*.md or docs/cli/commands.md.
  • Navigation is declared in mkdocs.yml under nav:. New top-level docs need an entry there.

Python tooling is required only for local preview and the strict build. If it is unavailable, report that the docs build was not checked; do not treat another green check as equivalent evidence.

Opening The PR

Make the PR easy to review. Include:

  • what changed
  • why the existing behavior or docs were not enough
  • what checks you ran locally
  • any follow-up work you intentionally did not include

Good PR titles:

  • docs: clarify first skill contribution path
  • feat: add <skill-name> skill
  • fix: tighten codex lifecycle guidance

Review Expectations

Maintainers will look for:

  • current frontmatter and taxonomy usage
  • linked references and working docs paths
  • matching runtime/story across docs and shipped artifacts
  • validation evidence
  • no secrets, symlinks, or dangerous shell patterns

Release Timing

AgentOps ships when the repo state justifies it. There is no fixed cadence.

Maintainer notes:

  • Keep [Unreleased] in CHANGELOG.md current.
  • Prefer coherent release bundles over random patch piles.
  • Draft releases are acceptable for validating packaging before public promotion.

Code of Conduct

Our Standards

Positive behavior:

  • Be respectful and inclusive
  • Provide constructive feedback
  • Collaborate openly
  • Welcome newcomers
  • Share knowledge generously

Unacceptable behavior:

  • Harassment or discrimination
  • Trolling or insulting comments
  • Personal or political attacks
  • Publishing others' private information
  • Other unprofessional conduct

Enforcement

Violations may result in:

  1. Warning from maintainers
  2. Temporary ban from contributing
  3. Permanent ban from project

Report issues to: fullerbt@users.noreply.github.com

Getting Help

Useful places to orient:

For examples, browse existing skills under skills/.

License

By contributing to this project, you agree that your contributions will be licensed under the Apache License 2.0.


Questions? Open an issue or email fullerbt@users.noreply.github.com