
Claude Code, Codex & Google Antigravity skills and hooks that make coding agents prove their work
Claude Code, Codex & Google Antigravity skills and hooks that make coding agents prove their work: risk-proportional workflows, CI proof gates, debugging, code review and a guard for destructive commands. npx senioros
·
Skills and hooks that make coding agents fix the cause, keep your history safe and prove the work.
Your agent loads the right skill on its own, asks before a destructive command, and cannot say "done" until your checks pass.
Measured on Claude Opus, Grok and Gemini against the model alone and three other skill packs.
Results · Quick start · How it works · Vibe coders · Senior developers · Skills · Proof · Evals · Architecture
A live bench runs the same model on the same tasks with no pack, with Superpowers, mattpocock/skills, compound-engineering and with SeniorOS. Hidden tests and the repository state grade every run; no model judges another.
| Model | No pack | Superpowers | mattpocock/skills | compound-engineering | SeniorOS |
|---|---|---|---|---|---|
| Claude Opus: a reported bug whose cause is in another module (3 held-out tasks, 5 trials) | 0/15 | 0/15 | 0/15 | 0/15 | 15/15 |
| Claude Sonnet 5.5: the same kind of bug (4 tasks, 5 trials) | 0/20 | 0/20 | not measured | not measured | 12/20 |
| Claude Opus: the same kind of bug, a fifth blind round (3 tasks, 5 trials) | 2/15 | 5/15 | 5/15 | 6/15 | 10/15 |
| Claude Opus: concurrency, flaky tests, stored-data changes, dependency upgrades (8 tasks, 3 trials) | 23/24 | not measured | not measured | not measured | 24/24 |
| Claude Opus: discipline (8 core tasks, 3 trials) | 21/24 | 19/24 | 21/24 | 20/24 | 24/24 |
| Grok 4.7, Grok CLI (13 tasks, 2 trials) | 15/26 | 14/26 | not measured | not measured | 23/26 |
| Gemini Pro, Antigravity (13 tasks, 2 trials) | 16/26 | 16/26 | not measured | not measured | 21/26 |

Watch the 60-second film: the same model on the same bug, alone and with SeniorOS, from two recorded runs.
main, Opus alone did it in one run of three. Grok, alone or with Superpowers, did it in every run. With SeniorOS the agent stopped and asked.Earlier results on smaller models:
| Model | No pack | Superpowers | SeniorOS |
|---|---|---|---|
| Claude Sonnet, 8 core tasks, 3 trials | 17/24 | 19/24 | 24/24 |
| Claude Haiku, 19 tasks, 3 trials | 52/57 | 53/57 | 57/57 |
Every run, task and caveat (including which tasks were used to tune the pack) is in evals/bench/, and you can rerun it with one command.
Measure your own pack. The bench takes any Claude Code plugin as an arm (--packs mine=you/your-pack@<commit>), and the leaderboard shows every table with its interval, its test against the model alone, its cost per pass and the commit of each pack. SUBMITTING.md has the commands.
Coding agents can already generate a lot of code. The expensive failures usually happen around the code:
SeniorOS treats these as engineering-system problems, not prompt-writing problems.
The goal is not to make the model sound senior. The goal is to make the workflow behave like disciplined engineering.
SeniorOS pushes agent work toward a risk-adaptive loop:
Goal Contract
↓
Relevant Context
↓
Risk Assessment
↓
Right-Sized Execution
↓
Proof
↓
Review
↓
Ship / Not Proven / BlockedThe important part is that not every task gets the same process.
| Task | SeniorOS behavior |
|---|---|
| Rename a button | Direct edit + proportionate verification |
| Add an API route | Light plan + relevant tests |
| Change authentication | Full context + security/risk review + proof |
| Production migration | Strict plan + rollback thinking + explicit boundaries |
| “Just make CI green” | Investigate root cause before weakening meaningful checks |
normal agent loop: request → generate → edit → "done"
SeniorOS loop: request → understand → assess → execute → prove → review → state what is knownSeniorOS prefers:
PROOF BASELINE: PASS
Residual risk: LOW
Skipped checks: noneinstead of:
Everything looks good! 🚀because optimism remains stubbornly absent from every serious test framework.
Requires Node.js 20 or newer. No clone needed: the CLI is published on npm as senioros, and its command is senior.
Claude Code (the plugin brings the skills, the routing rules, the destructive-command guard and the finish gate):
/plugin marketplace add joshsegatt/SeniorOS
/plugin install senioros@seniorosThat is all: the next session loads the right skill on its own. Other agents:
After installing, open /hooks in Codex and trust the three SeniorOS hooks (adapters/codex):
codex plugin marketplace add joshsegatt/SeniorOS
codex plugin add senioros@seniorosUntil it is listed in the Cursor marketplace, add a clone of this repository as a local plugin
(adapters/cursor).
Add it to opencode.json and restart OpenCode (adapters/opencode):
{ "plugin": ["senioros"] }Skills, routing rules and hooks in ~/.gemini, backed up and reversible:
npx senioros install --target antigravityOr as a native Antigravity plugin, with the skills and routing rules only (no guard or finish gate):
agy plugin install https://github.com/joshsegatt/SeniorOSUse one or the other, not both, or the skills are listed twice.
npx senioros install --target grokIt copies the plugin to ~/.grok/plugins/senioros/, enables it in ~/.grok/config.toml and adds the Grok routing rules
to ~/.grok/AGENTS.md (each file backed up once). See Grok CLI.
Run inside your project:
npx senioros install --target codex --scope project
npx senioros install --target claude --scope project # skills only, without the plugin's rules and guard
npx senioros install --target generic --scope project # portable Agent Skills layoutSeniorOS refuses to overwrite an existing skill unless you explicitly pass --force. In the bench, Claude never loaded
the skills on its own without the plugin's session-start rules, so prefer the plugin in Claude Code.
Upgrading from 0.1: skills are now prefixed with senior- (senior-debugging, senior-finish, …). senior install lists any old, unprefixed SeniorOS skills it finds so you can delete them; it never deletes them for you.
Run it once with npx, or install the senior command globally:
npm install -g seniorosThen, inside your project:
senior init # create the .senior/ project state
senior inspect # detected stack facts
senior risk --base main # risk of the current branch
senior finish # run the proof baselineEvery command also works without the global install, for example npx senioros risk --base main.
senior init creates a small, versionable project intelligence layer:
.senior/
├── README.md
├── project.json
├── decisions.md
└── work-ledger.jsonlgit clone https://github.com/joshsegatt/SeniorOS.git
cd SeniorOS
npm install
npm run check
npm link # puts your local build on PATH as `senior`
For vibe codersYou stay focused on what you want built. SeniorOS encourages the agent to handle the engineering beneath that request: What you gain
|
For senior developersSeniorOS gives the agent explicit engineering primitives instead of a persona prompt:
You keep architectural control. The harness reduces repeated supervision and makes agent behavior more inspectable. |
The foundation ships with six focused Agent Skills.
| senior-engineering Risk-adaptive routing and engineering discipline for non-trivial coding work. | senior-project-intelligence Builds and maintains concise, durable project context. | ||
| senior-debugging Reproduce, trace, test hypotheses, identify root cause, then fix minimally. | senior-code-review Evidence-based review for regressions, scope, security, and change quality. | ||
| senior-finish Final proof baseline and semantic completion gate. | senior-skill-authoring Creates focused SeniorOS-quality skills with eval contracts. |
Authoring rules for skills: How SeniorOS is built.
| The CLI | Every senior command, the proof order per language, regression proof, .senior/config.json |
| Host support | Claude Code, Codex, OpenCode, Antigravity, Grok CLI and the generic layout: what each install does |
| How SeniorOS is built | Proof Engine, project memory, evals, architecture, design principles |
| The bench | Every task, run and result behind the numbers above |
| Brand kit | Logos, the README images and how to regenerate them |
SeniorOS is not:
That last point matters. The project is shipping as a tested foundation rather than pretending its first release solved agentic engineering forever. Software already has enough heroic README fiction.
Next: richer goal and proof contracts, a selective context engine, host-specific eval runners and trace grading. See
ROADMAP.md.
Agent Skills can influence tool use and repository changes. Treat third-party skills as executable engineering policy, not harmless Markdown.
Read SECURITY.md.
SeniorOS should become stronger through evidence, not longer through accumulated folklore.
If a contribution adds process, it should answer:
Read CONTRIBUTING.md.
SeniorOS was created by Josh Segatt.
Building a portable engineering layer for coding agents around discipline, verification, and measurable progress rather than prompt theater.
Apache-2.0. See LICENSE.
Paste this prompt into your agent. It reads this page and does the setup for you.
Read https://aiagentslisting.com/mcp/senior-os to learn what the "Senior Os" MCP server does and how to install it. Add it to my coding agent's MCP configuration as documented on that page, then confirm the server connects and list the tools it exposes.Agents can also browse this directory over MCP at https://aiagentslisting.com/api/mcp
npx senioros install --target antigravity
npx senioros install --target grok
npx senioros install --target codex --scope project
npx senioros install --target claude --scope project # skills only, without the plugin's rules and guard
npx senioros install --target generic --scope project # portable Agent Skills layoutnpm install -g seniorosThis server runs locally, so we can't read its tool list over the web yet.
We could not run this server: its npm package has not answered in our offline sandbox, which usually means it needs an API key or other configuration to start. The four checks that need a live server are left out, and the grade tops out at B until we can read its tools.
Maintain this server? Claiming proves you own the listing and earns the last 4 points.
Claim this listingPaste this line near the top of your repository's README. The badge always renders the current score.
[](https://aiagentslisting.com/mcp/senior-os)Nothing comparable is listed yet.
v0.10.1SeniorOS 0.10.1Release notesv0.10.0SeniorOS 0.10.0Release notesv0.9.0SeniorOS 0.9.0Release notesv0.8.0SeniorOS 0.8.0Release notesv0.7.0SeniorOS 0.7.0Release notesv0.6.0SeniorOS 0.6.0Release notesv0.5.0Release notesv0.4.0SeniorOS 0.4.0Release notesv0.3.0SeniorOS 0.3.0Release notesv0.2.0SeniorOS 0.2.0Release notesConnect to the AI Agents Listing MCP server at https://aiagentslisting.com/api/mcp using the HTTP transport and complete the sign-in when it asks. Then call the submit_review tool for the listing "senior-os" with a rating from 1 to 5 and a short note on what worked and what did not.No reviews yet
Be the first to share your experience with this tool.
One email a week. New agents, MCP servers and skills, and what is actually getting traction.