tacticswitch
AGENTS.md0 starsChanged 3 months agoMITView on GitHub
TacticSwitch routes development work to the minimum sufficient agent topology: direct work for clear local tasks; independent contexts only when risk, uncertainty, or isolation justifies them. It layers above external workflows; it does not replace them. Python 3.11+. Distribution metadata: manifest.json. Portable core: templates/skills/; adapters: templates/adapters/; tests: tests/. templates/skills/ defines portable routing, role, packet, and acceptance contracts. templates/adapters/ translates those contracts for Codex, Claude, and Copilot. scripts/ installs and verifies package behavior. manifest.json is adapter-evidence authority. openspec/ remains framework-owned. Flow:…
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 32,671 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 repository16 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- more-with-lessSkill
- openspec-apply-changeSkill
- openspec-archive-changeSkill
- openspec-continue-changeSkill
- openspec-exploreSkill
- openspec-ff-changeSkill
- openspec-new-changeSkill
- openspec-onboardSkill
- openspec-proposeSkill
- openspec-sync-specsSkill
- openspec-update-changeSkill
- openspec-verify-changeSkill
- tactic-lean-orchestrateSkill
- tactic-spec-brainstormSkill
- tactic-spec-grillSkill
The file
We split this AGENTS.md into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are 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.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 33
Do not copy, edit, replace, or “simplify” canonical OpenSpec resources under .agents/skills/.
- Distinctive: line 34
Do not use subagents by default or ignore them categorically: choose the smallest route justified by the task.
- Very specific: line 35
Keep manifest.json claims, evidence, limitations, and adapter status coherent.
- Distinctive: line 37
Do not promote experimental adapter support from static checks alone.
- Very specific: line 48
Run relevant focused checks, then python scripts/quality_gate.py.
- Distinctive: line 54
Ask before changing public route semantics, adapter compatibility/support claims, installer scope or migration behavior, or canonical …
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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
Project Purpose
TacticSwitch routes development work to the minimum sufficient agent topology: direct work for clear local tasks; independent contexts only when risk, uncertainty, or isolation justifies them. It layers above external workflows; it does not replace them.
Stack and Runtime
Python 3.11+. Distribution metadata: manifest.json. Portable core: templates/skills/; adapters: templates/adapters/; tests: tests/.
Build, Test, Run
- Fast: focused
unittestmodule or relevant verification script. - Full:
python scripts/quality_gate.py. - High-risk/release assurance:
pythonscripts/adversarial_assurance.py all. - Install verification:
pythonscripts/verify_install.py --scope project --project /path/to/repository.
Architecture Map
templates/skills/ defines portable routing, role, packet, and acceptance contracts. templates/adapters/ translates those contracts for Codex, Claude, and Copilot. scripts/ installs and verifies package behavior. manifest.json is adapter-evidence authority. openspec/ remains framework-owned.
Flow: request → route decision → direct work or minimum independent contexts → one writer → fresh verification when required.
Domain Model
Routes: R0 direct; R1 primary plus fresh Verifier; R2 Scout discovery plus primary plus Verifier; R3 delegated Implementer plus Verifier; R4 isolated parallel Implementers plus integration verification.
A fresh Verifier did not implement the batch. One active writer owns overlapping files. Required isolation unavailable → BLOCKED; never silently substitute self-review.
The primary invokes canonical $openspec-apply-change for R0-R2 and exclusively owns the shared ledger and canonical apply for same-change R4; only an explicit R3 delegated Implementer boundary may invoke it.
Agent Guardrails
- Preserve external framework authority.
TacticSwitchis a layer above OpenSpec, TDD, and task systems. - Do not copy, edit, replace, or “simplify” canonical OpenSpec resources under
.agents/skills/. - Do not use subagents by default or ignore them categorically: choose the smallest route justified by the task.
- Keep
manifest.jsonclaims, evidence, limitations, and adapter status coherent. - Preserve one-writer ownership and fail-closed behavior where independent context is required.
- Do not promote experimental adapter support from static checks alone.
Known Failure Modes
- Treating all work as direct ignores material uncertainty or required independence.
- Spawning agents for every task adds cost, latency, and conflicting ownership.
- Claiming native multi-CLI behavior without current execution evidence overstates support.
- Replacing OpenSpec lifecycle semantics with orchestration-layer artifacts violates ownership.
Verification Before Completion
Run relevant focused checks, then python scripts/quality_gate.py.
For releases, protocol changes, or other high-risk work, run the applicable python scripts/adversarial_assurance.py <mode> check. Regenerate release hashes with python scripts/checksums.py --write when release assets change. Inspect the diff for unrelated files, caches, secrets, and evidence overclaims.
Escalation - Ask the User When
Ask before changing public route semantics, adapter compatibility/support claims, installer scope or migration behavior, or canonical OpenSpec resources. Ask when required independent worker capability is unavailable or public evidence is incomplete.
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.

