Stop babysitting your AI coding agent.
The Agent Ops Kit is the D.O.E. framework — Directive / Orchestration / Execution — that turns Claude Code, Cursor or Codex from a clever intern into a reliable operator: it reads an SOP, runs a deterministic script, and writes what it learned back into the SOP.
Secure checkout by Stripe. Works with Claude Code (CLAUDE.md + skills), Cursor (.cursorrules), Codex / Gemini CLI (AGENTS.md).
# CLAUDE.md (excerpt) Layer 1 — Directive: SOPs in directives/*.md Layer 2 — Orchestration: you. Route, don't do. Layer 3 — Execution: scripts in executions/ 1. Fix the root cause, not the symptom. A retry/reset/reprocess is RECOVERY, never a fix. 5. Every bug fix gets a test AND a standard. 6. Repair → deploy → verify → THEN recover. Mandatory checks before touching: | Deploys | directives/deploy_service.md | | Bug fixes | directives/core/bug-fix-standard.md |
Why agents fail on real work
A model that's 90 % reliable per step is 59 % reliable over five steps. Real tasks are ten to fifty steps. If the agent executes everything itself — fetching, parsing, transforming, deploying — compounding error does the rest, and you end up watching it.
D.O.E. moves everything with a right answer into tested code and leaves the agent exactly one job: deciding what to run next. Then it makes the system self-annealing — every failure becomes a test, a rule, and a lint, so month three is dramatically better than week one.
What's inside (34 files, ready to drop into any repo)
CLAUDE.md that enforces the loop
The 3 layers, 6 operating principles, the self-annealing loop, a mandatory-directive-checks table, and a definition of done. Battle-tested running a real business platform.
4 core directives
Operating principles, self-annealing loop, bug-fix standard, agent governance — written to sync unchanged into every project you run.
3 complete example directives
Deploy a service, QA a website, scrape a site — each with inputs → tools → outputs → edge cases → verification, pre-loaded with the gotchas.
7 execution scripts
Script template (argparse, .env, retries, exit codes), directive↔script linter, secrets linter, anneal log, headless QA/SEO crawler, SSH deploy with CRLF/BOM fix, robots-aware scraper.
Tests that make the kit eat its own rules
12 tests for the linters, including "the kit itself passes". Every tool has a test — that's the standard.
Multi-agent templates
PM agent + worker agent definitions, an agent registry schema, work-authorization and meta-work-cap rules from real incidents.
3 Claude Code skills
qa-test, deploy, project-status — ten lines each, calling the scripts, so expensive reasoning happens once.
3 guides
Why D.O.E. works · Adopting it in an existing codebase (day 1 / week 1 / month 1) · Running a fleet of agents without burning a fortune.
File tree
agent-ops-kit-v1/
START-HERE.md
CLAUDE.md
LICENSE.txt
CHANGELOG.md
directives/
00_template.md
core/operating-principles.md
core/self-annealing-loop.md
core/bug-fix-standard.md
core/agent-governance.md
examples/deploy_service.md
examples/qa_website.md
examples/scrape_website.md executions/
00_template.py
lint_directives.py
lint_secrets.py
anneal.py
qa_website.py
deploy_service.py
scrape_website.py
tests/test_lint_directives.py
tests/test_lint_secrets.py
skills/qa-test · deploy · project-status
agents/pm-agent.md · worker-agent.md
registry.example.json
docs/why-doe.md · adopting-doe.md
running-a-fleet.mdTwo real findings from building this very store
A live Stripe secret turned up hard-coded in an old helper script in the seller's own repo. lint_secrets.py now runs pre-commit so it can't happen again — and lint_directives.py caught the kit's own CLAUDE.md pointing at a directive that didn't exist:
$ py executions/lint_secrets.py ERROR executions/create_links.py:3: stripe live secret lint_secrets: FAIL (1 findings) $ py executions/lint_directives.py ERROR CLAUDE.md: references directives/scrape_website.md which does not exist lint_directives: FAIL (7 directives, 1 errors)
(Both fixed before release. The kit follows its own rules: every tool has a test, and the kit itself must pass its linters.)
Who it's for
| You | What changes |
|---|---|
| Solo dev / founder shipping with Claude Code or Cursor | The agent reads an SOP before acting and stops re-learning your gotchas every session. |
| Small team adopting agents | Shared directives are your team's operating manual; new people (and new agents) onboard from the same files. |
| Running several agents | Work authorization, one-agent-per-capability, deploy approval gate, meta-work cap — rules that each cost a real incident to learn. |
FAQ
Does it only work with Claude Code?
No. CLAUDE.md is the canonical file; Cursor users paste it into .cursorrules, Codex and Gemini CLI users into AGENTS.md / GEMINI.md. The directives and scripts are tool-agnostic (plain Markdown + Python 3.10+).
Is there a free version?
Yes — a lite version (CLAUDE.md + one core directive + the secrets linter) is on GitHub. The paid kit adds the other directives, all scripts and tests, the skills, the agent templates and the guides.
How is it delivered?
Instantly: a download page right after checkout plus the same link by email (valid 30 days). Updates are emailed to every buyer free.
Refunds?
14 days, no questions — reply to your receipt email.
Who made this?
Calyx AI is a product line of Quality Tester Solutions LLC (St. Petersburg, FL), a small engineering company that builds test equipment and runs its own CRM, invoicing and inventory platform with a fleet of AI agents using exactly this framework.
Agent Ops Kit v1
Instant download · free updates · 14-day refund · license for unlimited projects you own or work on
Buy now