Continuous Agent Loop
Run autonomous agent loops with quality gates, evals, and recovery controls
- Repository
- affaan-m/ECC
- Stars
- 295,127
- Licence
- MIT
- Files
- 1 file
What it does
This skill teaches an AI agent a pattern for running long, self-driving work sessions that repeat steps until a task is done. It helps the agent pick the right loop style (sequential, PR-focused, RFC-decomposition, or exploratory parallel generation), suggests a production stack that combines quality checks and evaluations, and lists common failure modes such as looping without progress, repeated retries on the same cause, and runaway cost. When something goes wrong it recommends freezing the loop, running a harness audit, narrowing scope to the failing unit, and replaying with explicit acceptance criteria.
Great for
- Developers who want an AI agent to work through multi-step coding tasks with checkpoints
- Teams setting up automated PR or CI-driven agent workflows who want guardrails against cost drift
- People who already use the ECC skill collection and want a coordinating loop pattern
Skip it if
- Simple one-off questions or small edits, where a loop adds overhead without benefit
- Users without the referenced companion skills installed, since the pattern points to other skills (such as eval-harness and plankton-code-quality) that this file does not contain
- Unsupervised long-running loops on production systems without spending limits, because the file describes cost drift as a risk but does not itself enforce any limit
Try asking
- “Set up a sequential agent loop with quality gates for implementing a feature in this repo”
- “Choose the right loop style for splitting a large RFC into tasks”
- “Recover a stalled agent loop: freeze it, audit the harness, and narrow scope to the failing unit”
- “Plan a production agent stack with evaluation and session persistence steps”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
continuous-agent-loop.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip continuous-agent-loop.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip continuous-agent-loop.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Read the loop selection flow first and default to sequential unless you need CI control, RFC decomposition, or parallel exploration
- Set explicit acceptance criteria and a cost or iteration cap before starting any loop
- The file is written in Japanese; ask the agent to summarize or follow it in your preferred language if needed
What's inside
- SKILL.md
Redistributed unmodified from affaan-m/ECC under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.