Instructions file CodexOpenCode
AGENTS.md instructions for zhyx12/projtool, covering agents.md — for developers working on projtool, what projtool is, in one paragraph, critical: you are working on projtool, not with projtool, current state and the asset boundary.
Instructions file
Claude Code instructions for zhyx12/projtool, covering claude.md — for developers working on projtool, what projtool is, in one paragraph, critical: you are working on projtool, not with projtool, current state (updated 2026-07-22) and the asset boundary.
Skill Claude CodeCodex
Use this skill when diagnosing failures during a projtool image build — that is, when mcpprojtoolremoteexec returns a non-zero exit code while running setup.sh inside a builder instance, or when the user explicitly asks for help debugging an image build / container setup. Trigger on phrases like "image build failed"…
Skill Claude CodeCodex
Use this skill whenever the user is doing ML experiment work in a projtool-managed repo — that is, requests involving training, runs, manifests, reports, mutagen sync, AutoDL instances, GPU provisioning, experiment branches, worktrees, image builds, research threads / lineage / the docs registry, or projtool tool…
Command
Power off (or release) the AutoDL instance for the current experiment to stop billing.
Command
Start a training run on the current experiment branch. Provisions an instance if needed, syncs code, and launches in tmux.
Command
Show the status of all in-flight runs and instances for this project.
Command
Adopt a pre-registry docs layout (plans/, exp/) into the research-thread registry — scan for drift, dry-run the migration proposal, confirm the mapping WITH THE USER, then apply. Never guess history.
Command
Read the experiment record (journal, findings, reports, run table) and propose the next 1-3 experiments with rationale and cost — the user decides, then execution goes through startruns/starttraining.
Command
Start a new analysis report in the independent docs repo, prefilled with frontmatter from the cited runs (single branch or cross-branch summary).
Command
Distill this work session's experiment outcomes into the experiment's findings.md (what was tried, what failed and why, what's now established) and commit it to the docs repo.
Command
Close a research thread — conclude it with an evidence-citing conclusion.md, or abandon it with a one-line reason. Terminal is terminal; to revive, open a successor thread with a supersedes edge.
Command
Open a new research thread (a research line) in the docs registry — pick a slug, create the record + idea.md/plan.md skeletons, interview the user to fill them, and point forward to the first experiment.
Command
Babysit every active run to completion — staged polling, auto-finalize, progress report; no raw SSH loops.
Skill Claude CodeCodex
Use this skill ONLY when the user explicitly asks to set up, install, or onboard projtool in a project. Trigger on phrases like "I want to use projtool", "I just installed projtool", "set up projtool here", "retrofit this project with projtool", "接入 projtool", "用 projtool 管理这个项目", "改造项目使用 projtool". This skill is…
MCP server Claude CodeCodexCursor
MCP server "projtool" as configured in zhyx12/projtool. Runs locally from the projtool Python package.