Agent skill
harvest
Cross-rig knowledge consolidation. One-time sweep + ongoing tiered promotion. Walks all .agents/ directories, extracts learnings/patterns/research, deduplicates across rigs, and promotes high-value items to global hub. Triggers: "harvest", "consolidate knowledge", "cross-rig sweep", "knowledge federation", "harvest knowledge".
Install this agent skill to your Project
npx add-skill https://github.com/boshu2/agentops/tree/main/skills-codex/harvest
SKILL.md
Harvest — Cross-Rig Knowledge Consolidation
Nightly usage (2026-04-09):
$dream startnow runs harvest as part of its bounded compounding loop. Use$harvestfor manual sweeps, CI runs, or when Dream is disabled. Dream holds.agents/overnight/run.lockwhile running — manualao harvestwill refuse until the lock releases.
Sweep all .agents/ directories across the workspace, extract learnings, patterns,
and research, deduplicate cross-rig, and promote high-value items to the global
knowledge hub (~/.agents/learnings/).
What This Skill Does
The knowledge flywheel captures learnings per-rig, but they stay siloed. Harvest
closes the loop by walking all rigs, extracting artifacts, deduplicating by content
hash, and promoting high-confidence items to the global hub where every rig can
access them via ao inject.
When to use: Before an evolve cycle, after a burst of development across multiple rigs, or weekly as part of knowledge governance.
Output: .agents/harvest/latest.json (catalog) + promoted files in ~/.agents/learnings/
Execution Steps
Step 1: Preview Scope (Dry Run)
ao harvest --dry-run --quiet
Read .agents/harvest/latest.json and report:
- Rigs discovered
- Total artifacts extracted
- Unique vs duplicate count
- Promotion candidates (artifacts >= min confidence)
Step 2: Confirm Execution
Skip if --auto is set. Otherwise, show the dry-run summary and ask:
Harvest will promote N artifacts from M rigs to ~/.agents/learnings/.
Proceed? [Approve / Adjust threshold / Abort]
Step 3: Execute Harvest
ao harvest --roots ~/gt/ --promote-to ~/.agents/learnings --min-confidence 0.5
Step 4: Post-Harvest Cleanup
Run dedup on the promotion target to clean up any remaining duplicates:
ao dedup --merge ~/.agents/learnings/ 2>/dev/null || true
Step 5: Report Results
Report to user:
- Rigs scanned
- Artifacts extracted and unique count
- Duplicates found (with top duplicate groups)
- Artifacts promoted (with provenance)
- Top discoveries (highest-confidence cross-rig patterns)
Flags
| Flag | Default | Description |
|---|---|---|
--auto |
off | Skip confirmation gate |
--roots |
~/gt/ |
Override root directories to scan |
--min-confidence |
0.5 | Minimum confidence for promotion |
--include |
learnings,patterns,research |
Artifact types to extract |
Quick Start
$harvest # Full sweep with confirmation
$harvest --auto # Hands-free sweep
$harvest --min-confidence 0.7 # Only promote high-confidence items
$harvest --roots ~/gt/,~/projects/ # Scan additional directories
Governance
See references/governance.md for ongoing governance model: size budgets, sweep frequency, staleness thresholds, and cross-rig synthesis triggers.
See Also
$compile— Single-rig Mine/Grow/Defrag$flywheel— Flywheel health monitoring$inject— Knowledge injection into sessions$forge— Transcript knowledge extraction
Reference Documents
- references/governance.md — Governance model for ongoing knowledge consolidation
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
swarm
Spawn isolated Codex sub-agents for parallel task execution using the current runtime primitives. Triggers: "swarm", "spawn agents", "parallel work", "run in parallel", "parallel execution".
council
Multi-perspective review for Codex using the current sub-agent runtime. Triggers: "council", "get consensus", "multi-model review", "multi-perspective review", "council validate", "council brainstorm", "council research".
openai-docs
Use when the user asks how to build with OpenAI products or APIs and needs up-to-date official documentation with citations (for example: Codex, Responses API, Chat Completions, Apps SDK, Agents SDK, Realtime, model capabilities or limits); prioritize OpenAI docs MCP tools and restrict any fallback browsing to official OpenAI domains.
crank
Hands-free epic execution for Codex using wave-based sub-agents and lead-side validation. Triggers: "crank", "run epic", "execute epic", "run all tasks", "hands-free execution", "crank it".
pr-retro
Learn from PR outcomes. Analyzes accept/reject patterns and updates contribution lessons. Triggers: "pr retro", "learn from PR", "PR outcome", "why was PR rejected", "analyze PR feedback".
ratchet
Brownian Ratchet progress gates for RPI workflow. Check, record, verify. Triggers: "check gate", "verify progress", "ratchet status".
Didn't find tool you were looking for?