Autoloop Controller
v1.1.1When continuous automated improvement of a Skill is needed. Wraps improvement-orchestrator in a persistent loop with convergence detection (plateau/oscillati...
⭐ 0· 75·1 current·1 all-time
by_silhouette@lanyasheng
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (persistent loop around an improvement orchestrator) matches the implementation: autoloop.py invokes an on-repo improvement-orchestrator subprocess, persists per-iteration state and logs, and performs convergence/cost checks. Declared dependencies (Python, repo-local lib.common, improvement-orchestrator) are reasonable and proportional to the stated purpose.
Instruction Scope
SKILL.md and scripts instruct the agent to run the orchestrator subprocess, persist state and iteration logs to disk, and optionally schedule runs via cron. These actions are within the described scope. Note: the controller executes another on-repo script (improvement-orchestrator) and reads/writes files under the provided state-root; it does not independently request credentials or contact external endpoints in the provided code.
Install Mechanism
No install spec or remote downloads are present — this is an instruction/code bundle that runs with the local Python environment. That is lowest-risk for installation; nothing is fetched from external URLs.
Credentials
The skill requires no environment variables or external credentials. It interacts with filesystem paths you provide (target skill path and state-root) and expects a local improvement-orchestrator. No unrelated secrets or cloud credentials are requested.
Persistence & Privilege
The skill persists its own state and logs to disk and is designed to be run from cron/long-running sessions. always:false (normal). Persistence is expected for resuming long-running improvement loops; however, scheduled/cron usage means the code will run on a schedule and write logs/state to the specified directory—ensure you choose the state path and schedule carefully.
Assessment
This skill appears to do what it says: repeatedly run a local improvement-orchestrator and persist per-iteration state and logs. Before installing or scheduling it, verify the following: 1) Confirm the improvement-orchestrator script it invokes (./skills/improvement-orchestrator/scripts/orchestrate.py) is present and trustworthy — the autoloop runs that subprocess and will inherit whatever that script does. 2) Choose an explicit state-root and log location you control (avoid leaving defaults in world-writable /tmp if you care about confidentiality), and inspect saved files (autoloop_state.json, iteration_log.jsonl) for any sensitive outputs. 3) Set conservative cost caps (--max-cost) and max-iterations before running unattended to avoid unexpected compute/LLM costs. 4) If you plan to use cron, review run-eval.sh and the crontab entry to ensure scheduling and file permissions meet your policies. 5) The controller imports repo-local lib.common by manipulating sys.path — make sure the repository root you install under is the intended codebase. If any of these are unclear, inspect the improvement-orchestrator code and run the controller in dry-run (--dry-run or equivalent) / single-run mode first.Like a lobster shell, security has layers — review code before you run it.
latestvk974md7fjre4jdpqv4a2vaqvbh84c3gp
License
MIT-0
Free to use, modify, and redistribute. No attribution required.