Deletes torch/ (5947 tracked files, 86M) and everything that existed only to
wire it up: torch-config.json, the root package.json and package-lock.json
-- whose sole dependency was torch-lock and whose six scripts were all
torch:* -- and the TORCH Memory Integration sections of CLAUDE.md and
AGENTS.md.
Nothing functional referenced it. No CI workflow mentions torch, no Python
or JS source imports torch-lock, and the only remaining occurrence outside
historical logs is a stray comment in perf/perf_agent.py. The determinism
gate is green after removal.
Two entropy-audit findings disappear with it rather than needing fixes: the
predictable temp filename and the Math.random value misnamed `nonce` (L4,
already fixed in 7047a5d but now moot), and the 18 test failures that bare
`node --test` produced by discovering stale code under torch/_backups/.
Kept deliberately: .scheduler-memory/, memory-update.md and memory-updates/.
Those are accumulated notes rather than tooling, and deleting someone's
notes is a different decision from deleting the tool that wrote them. They
can go separately if wanted.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01HtNMsKzzFTPURemX7tT19u
3.9 KiB
SeedPass Agent Guide
Overstory Ecosystem
This project is part of the Overstory multi-agent development ecosystem. When asked to "update Overstory" or work on the ecosystem, these are the relevant repositories:
- Overstory — The coordinator. Orchestrates multi-agent workflows across worktrees.
- Mulch — Structured expertise/memory system for agents.
- Canopy — Prompt management system for agents.
- Seeds — Git-native issue tracker for agents.
- os-eco — The entire Overstory ecosystem (meta-repo/docs).
Planning & Strategy
Consult these first to align with current objectives:
- Dev Control Center: Current priorities.
- TUI v2 Execution Plan: Roadmap for TUI v2.
Testing
Always run the full suite before proposing changes:
bash scripts/run_ci_tests.sh- Central Index: Test Infrastructure Inventory
Project Expertise (Mulch)
This project uses Mulch for structured expertise management.
At the start of every session, run:
mulch prime
This injects project-specific conventions, patterns, decisions, and other learnings into your context.
Use mulch prime --files src/foo.ts to load only records relevant to specific files.
Before completing your task, review your work for insights worth preserving — conventions discovered, patterns applied, failures encountered, or decisions made — and record them:
mulch record <domain> --type <convention|pattern|failure|decision|reference|guide> --description "..."
Link evidence when available: --evidence-commit <sha>, --evidence-bead <id>
Run mulch status to check domain health and entry counts.
Run mulch --help for full usage.
Mulch write commands use file locking and atomic writes — multiple agents can safely record to the same domain concurrently.
Before You Finish
- Discover what to record:
mulch learn - Store insights from this work session:
mulch record <domain> --type <convention|pattern|failure|decision|reference|guide> --description "..." - Validate and commit:
mulch sync
Issue Tracking (Seeds)
This project uses Seeds for git-native issue tracking.
At the start of every session, run:
sd prime
This injects session context: rules, command reference, and workflows.
Quick reference:
sd ready— Find unblocked worksd create --title "..." --type task --priority 2— Create issuesd update <id> --status in_progress— Claim worksd close <id>— Complete worksd sync— Sync with git (run before pushing)
Before You Finish
- Close completed issues:
sd close <id> - File issues for remaining work:
sd create --title "..." - Sync and push:
sd sync && git push
Prompt Management (Canopy)
This project uses Canopy for git-native prompt management.
At the start of every session, run:
cn prime
This injects prompt workflow context: commands, conventions, and common workflows.
Quick reference:
cn list— List all promptscn render <name>— View rendered prompt (resolves inheritance)cn emit --all— Render prompts to filescn update <name>— Update a prompt (creates new version)cn sync— Stage and commit .canopy/ changes
Do not manually edit emitted files. Use cn update to modify prompts, then cn emit to regenerate.