Continuous Agent Loop
Run long-running autonomous agent work with explicit loop choice, quality gates, evals, and recovery when progress stalls or cost drifts.
Overview
Continuous-agent-loop is a journey-wide agent skill that patterns autonomous loops with quality gates, evals, and recovery—usable whenever a solo builder needs controlled long-running agent work before merging or shippin
Install
npx skills add https://github.com/affaan-m/everything-claude-code --skill continuous-agent-loopWhat is this skill?
- Loop selection flow: continuous-pr for strict CI/PR control, rfc-dag for RFC decomposition, infinite for exploratory par
- Recommended production stack chains RFC decomposition, plankton-code-quality with /quality-gate, eval-harness, and nanoc
- Documents failure modes: loop churn, repeated retries, merge queue stalls, and unbounded cost escalation
- Recovery playbook: freeze loop, run /harness-audit, narrow scope to failing unit, replay with explicit acceptance criter
- Canonical v1.8+ name superseding autonomous-loops for compatibility
- 4-mode loop selection flow
- 4-step recommended production stack (RFC, quality gate, eval, session persistence)
- 4 documented failure modes
Adoption & trust: 4.5k installs on skills.sh; 210k GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
Your agent keeps looping without measurable progress, burning tokens on the same failure while PRs and merge queues stall.
Who is it for?
Indie builders running ECC-style agent stacks who need RFC decomposition, PR-bound loops, or parallel exploration with explicit stop conditions.
Skip if: One-shot chat tasks or manual edits where no harness, acceptance criteria, or CI gate exists—start simpler before enabling infinite or continuous-pr modes.
When should I use this skill?
Patterns for continuous autonomous agent loops with quality gates, evals, and recovery controls; superseding autonomous-loops as the v1.8+ canonical loop skill.
What do I get? / Deliverables
You pick an appropriate loop mode, stack gates and evals, and follow a freeze-audit-narrow-replay recovery path when churn or cost drift appears.
- Chosen loop mode and rationale
- Stacked gate-and-eval configuration
- Recovery runbook applied to failing scope
Recommended Skills
Journey fit
Useful at every journey phase - explore requirements and options before committing to a direction.
Where it fits
Select sequential vs rfc-dag before decomposing a large feature into agent-sized tasks.
Run continuous-pr with plankton-code-quality and /quality-gate before merging agent-generated diffs.
Freeze an infinite exploration loop when merge queue stalls and replay one failing unit with acceptance criteria.
Invoke /harness-audit after repeated retries share the same root cause.
Pair nanoclaw-repl persistence with eval-harness so sessions survive long RFC pipeline runs.
How it compares
Use as the orchestration layer over ad-hoc 'keep going' prompts—not a substitute for eval-harness or quality-gate skills it assumes in the production stack.
Common Questions / FAQ
Who is continuous-agent-loop for?
Solo builders and small teams using Claude Code-style stacks who run multi-step autonomous work and need loop selection plus recovery discipline.
When should I use continuous-agent-loop?
In Build when wiring agent-tooling; in Ship when continuous-pr and quality gates guard merges; in Operate when debugging loop churn; anytime you need RFC-dag or infinite parallel modes with eval recovery.
Is continuous-agent-loop safe to install?
Autonomous loops can modify repos and trigger CI—review Security Audits on this page and cap scope before enabling infinite or continuous-pr patterns.
SKILL.md
READMESKILL.md - Continuous Agent Loop
# Continuous Agent Loop This is the v1.8+ canonical loop skill name. It supersedes `autonomous-loops` while keeping compatibility for one release. ## Loop Selection Flow ```text Start | +-- Need strict CI/PR control? -- yes --> continuous-pr | +-- Need RFC decomposition? -- yes --> rfc-dag | +-- Need exploratory parallel generation? -- yes --> infinite | +-- default --> sequential ``` ## Combined Pattern Recommended production stack: 1. RFC decomposition (`ralphinho-rfc-pipeline`) 2. quality gates (`plankton-code-quality` + `/quality-gate`) 3. eval loop (`eval-harness`) 4. session persistence (`nanoclaw-repl`) ## Failure Modes - loop churn without measurable progress - repeated retries with same root cause - merge queue stalls - cost drift from unbounded escalation ## Recovery - freeze loop - run `/harness-audit` - reduce scope to failing unit - replay with explicit acceptance criteria