torque
AGENTS.md0 starsChanged 3 days agoMITView on GitHub
This file is the single source of instructions for any coding agent working on this repository (Claude Code, Codex, Gemini CLI, Cursor or others). Tool-specific files only point here. Torque is a Python command-line tool plus packaged workflows that give a coding assistant a private, local workspace per Salesforce client or internal initiative. This repository is the public, generic engine. Private firm, client and personal material belongs outside it. Necessary public attribution, license notices and dependency provenance are retained. All…
- Installs packages
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 33,315 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository9 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- salesforce-architecture-reviewSkill
- salesforce-buildSkill
- salesforce-code-analyzer-reviewSkill
- salesforce-deliverySkill
- salesforce-flowSkill
- salesforce-nonprofit-cloudSkill
- salesforce-npspSkill
- salesforce-soql-reviewSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 52
All of the commands above pass locally, the change has tests, CI is green on Linux, macOS and Windows, and user-facing behavior changes are …
- Distinctive: line 59
Use the user's existing Salesforce access and authorization.
- Distinctive: line 61
Prefer integrating a mature existing tool over building one.
- Distinctive: line 63
No commit, push or release is implied by a local change.
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Torque: instructions for coding agents
This file is the single source of instructions for any coding agent working on this repository (Claude Code, Codex, Gemini CLI, Cursor or others). Tool-specific files only point here.
What this repo is
Torque is a Python command-line tool plus packaged workflows that give a coding assistant a private, local workspace per Salesforce client or internal initiative. This repository is the public, generic engine. Private firm, client and personal material belongs outside it. Necessary public attribution, license notices and dependency provenance are retained.
Map
| Path | Contents |
src/torque/ | The engine: workspace and records, change records, gate and connected mode, approvals, upgrades, demo |
src/torque/data/ | Generated bundle that torque workspace init and upgrade install into a workspace (do not edit by hand) |
packages/ | Bundled Salesforce tools reached through torque (revert, qa, browser, logs, advisory, lesson, meeting); provenance in packages/provenance.json |
workflows/ | Source recipes (*.md), catalogue (catalogue.json) and adapter generator (sync_adapters.py) |
.claude/commands/ | Generated workflow adapters; edit workflows/*.md, then regenerate |
.claude/rules/, .claude/agents/, .agents/skills/ | Source standing rules, worker roles and skills; bundled with the adapters into src/torque/data/ |
lessons/ | Source platform lessons, one Markdown file each (format: docs/lessons.md); bundled into src/torque/data/lessons/ |
floor/AGENTS-floor.md | Source of the Torque block every host loads from a workspace's AGENTS.md (at most 60 lines; docs/workspace-upgrades.md); bundled into src/torque/data/floor/ |
tests/, scripts/ | Offline tests and release checks |
docs/, examples/ | User guides, one validation record per release, design records |
Agent assets, by tool
- Workflow recipes:
torque workflows listandtorque workflows show NAMEwork for any agent. Claude Code also reads them as.claude/commands/. - Skills:
.agents/skills/(read by agents that support the shared skills folder); workspaces get the same skills in.claude/skills/. List:docs/skills.md. - Standing rules:
.claude/rules/*.mdare plain Markdown; any agent should read them before client work. - Worker roles:
.claude/agents/*.mdare plain role descriptions any agent can follow. - Antigravity (
agy) reads.agents/only. A workspace gets the rules, recipes (as skills and slash commands) and worker roles there in its format, derived atinitandupgradebysrc/torque/antigravity.py; nothing for it is generated in this repository. Seedocs/workspace-upgrades.md. - Guardrails: the gate (
src/torque/gate.py) runs as a Claude Code hook;src/torque/gate_antigravity.pyruns the same gate as an Antigravity hook.src/torque/hosts.pyis the registry of hosts:torque launchstarts the host a workspace names, or--host claude|antigravity. Connected mode under Antigravity is tested offline only. Other agents get the same records and workflows but not the hook. Seedocs/ai-access.mdanddocs/hosts.md.
Commands
python -m pip install -e '.[dev]'
python scripts/test-offline.py -q # full offline suite and package self-tests
python workflows/sync_adapters.py --check # catalogue matches the recipe files
python scripts/sync-workflows.py --check # src/torque/data matches the sources
python scripts/check-provenance.py # carried-over package files match their records
python -m build && python scripts/check-distribution.py dist/*.whl --sdist dist/*.tar.gz
python scripts/smoke-installed.py --require-wheel # run from a clean venv with the wheel installedDone means
All of the commands above pass locally, the change has tests, CI is green on Linux, macOS and
Windows, and user-facing behavior changes are reflected in README.md, the relevant docs/
page and CHANGELOG.md. Report observed, asserted, unknown and untested outcomes separately.
Rules
- No private firm, client or personal material, credentials, org mappings or real record data in tracked files. Necessary public authorship, LICENSE/NOTICE text, dependency attribution and provenance are allowed; retain required notices. Fictional fixtures use reserved example domains. See
CONTRIBUTING.md;tests/test_public_hygiene.pyenforces part of this policy. - Use the user's existing Salesforce access and authorization; never select another client or org silently. Live org and browser tests run only when explicitly in scope.
- Edit recipes in
workflows/*.mdand the catalogue inworkflows/catalogue.json; runpython workflows/sync_adapters.pyto regenerate.claude/commands/, thenpython scripts/sync-workflows.pyto rebuildsrc/torque/data/. Edit standing rules in.claude/rules/, worker roles in.claude/agents/, skills in.agents/skills/, platform lessons inlessons/and the workspace block infloor/AGENTS-floor.md(raise the version in its first line when its text changes), then rebuild the bundle. Do not edit generated adapters or bundled files by hand. - Prefer integrating a mature existing tool over building one; no feature-count targets or extra user ceremonies.
- Retain material adoption decisions in
docs/research-adoption.md. - No commit, push or release is implied by a local change; the maintainer decides.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

