edca3582d5
All checks were successful
CI / validate (push) Successful in 1m24s
Per saved memory feedback_act_runner_needs_config_file_env: runners 1-8 were spawned without -e CONFIG_FILE=/config.yaml; act_runner fell back to /data/config.yaml and ignored runner.envs the whole time. Orchestrator recreated 1-8 with full proper env. All 16 now uniform with AGENT_TOOLSDIRECTORY + RUNNER_TOOL_CACHE + GITHUB_SERVER_URL + GH_HOST. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .github/workflows | ||
| .molecule-ci/scripts | ||
| adapters | ||
| skills/cron-learnings | ||
| .gitignore | ||
| CLAUDE.md | ||
| known-issues.md | ||
| plugin.yaml | ||
| README.md | ||
molecule-skill-cron-learnings — Per-Tick Operational Memory
Plugin for Claude Code and Hermes. End each cron tick by appending 1–3 lines of operational learnings (what worked, what surprised, what to change next tick) to a per-project JSONL file. Replay at the start of the next tick so the agent carries operational context across session boundaries.
Pairs with molecule-session-context for auto-loading learnings at session start.
What it does
At the end of each cron tick:
- Review what happened — what worked, what failed, what was unexpected
- Append 1–3 lines to
~/.claude/projects/<project>/cron-learnings.jsonl - Next tick start: replay learnings so the agent knows what changed
JSONL format
{"tick": "2026-04-21T14:00Z", "role": "research-lead", "learnings": ["GH_TOKEN was 401 for 3h — token refresh is fragile", "PR template needs a 'Testing' section"]}
{"tick": "2026-04-21T15:00Z", "role": "sdk-lead", "learnings": ["a2a-sdk 1.0.0 broke workspace startup — pinned to 0.3.x"]}
Installation
In org template (org.yaml)
plugins:
- molecule-skill-cron-learnings
From URL
github://Molecule-AI/molecule-ai-plugin-molecule-skill-cron-learnings
License
Business Source License 1.1 — © Molecule AI.