open-ps3005d
AGENTS.md1 starsChanged 21 days agoMITView on GitHub
Working charter for AI coding agents on this repository (CLAUDE.md maps this file for Claude Code). Bootstrap Rule (Existing or Empty Repo): When .agents/init_memory.sh has not been run or Section 1 still contains TODO: placeholders: 1. Run bash .agents/init_memory.sh first (creates .agents/memory/*, updates .gitignore, and links .claude/). 2. If existing files are present (pyproject.toml, package.json, Cargo.toml, CMakeLists.txt, Makefile, README.md, etc.), inspect them and replace the TODO: placeholders below automatically. 3. If this is a brand-new / empty repository, ask the…
- Reads credentials
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 repository2 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- team-orchestrationSkill
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 2 are common, 4 are distinctive, and 5 are very specific to this project.
≈2
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 24
Hardware safety: never exceed 30.00 V / 5.000 A.
- Distinctive: line 25
Output is off by default: never switch the output on implicitly (on connect, on recall, on setpoint change).
- Distinctive: line 26
Single serial owner: only one process may hold the port.
- Very specific: line 27
Protocol timing: no terminators; the device garbles replies when paced too fast — keep ≥100 ms after writes and ≥40 ms after queries …
- Distinctive: line 29
Front-panel memories are user data: never overwrite M1–M5 without the user's consent.
- Very specific: line 30
Must run on Windows and Linux: no POSIX-only APIs.
- Distinctive: line 39
Commit only when explicitly asked by the user.
- Very specific: line 40
Never commit .agents/memory/, build artifacts, local captures, or temporary debug files.
- Common: line 46
Zero Secrets in Git: NEVER hardcode, stage, or commit secrets, API keys, tokens, passwords, private keys, or credentials into Git.
- Very specific: line 48
Agent Memories & Docs: NEVER write raw secrets, tokens, or credentials into .agents/memory/*.md, artifacts, logs, or documentation.
- Common: line 49
Pre-Commit / Review Gate: Always verify git status and git diff contain no secrets or unintended files before finishing or approving any …
≈ “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
AGENTS.md — Repository Working Charter
Working charter for AI coding agents on this repository (
CLAUDE.mdmaps this file for Claude Code).Bootstrap Rule (Existing or Empty Repo): When
.agents/init_memory.shhas not been run or Section 1 still containsTODO:placeholders:
- Run
bash .agents/init_memory.shfirst (creates.agents/memory/*, updates.gitignore, and links.claude/).- If existing files are present (
pyproject.toml,package.json,Cargo.toml,CMakeLists.txt,Makefile,README.md, etc.), inspect them and replace theTODO:placeholders below automatically.- If this is a brand-new / empty repository, ask the user what we are building and which tech stack to use (or use the description in their first prompt), scaffold the project, and fill in the
TODO:placeholders below and.agents/memory/project_state.md.
1. Project Configuration & Quality Gates (All Agents)
- Project Overview:
- Name & Purpose:
ps3005d— cross-platform (Windows + Linux) control of the Velleman PS3005D 0–30 V / 0–5 A lab power supply over USB/RS232: set V/I, output/OVP/OCP, memory M1–M5, live graphs/logging, plus an MCP server so AI agents can drive the supply during hardware testing. - Tech Stack: Python ≥ 3.11,
pyserial(transport), PySide6 +pyqtgraph(GUI, extragui), officialmcpPython SDK / FastMCP over stdio (extramcp), hatchling build, venv in.venv/. - Entrypoints / Packages:
src/ps3005d/—protocol.py(pure command build/parse, no I/O),transport.py(paced serial I/O),device.py(PowerSupplydriver),sim.py(simulated supply),controller.py(limits + locking for remote use),mcp_server.py(ps3005d-mcp),gui/(ps3005d-gui, PySide6 worker thread + main window),cli.py(ps3005d). Tests intests/.PC_SOFTWARE/is the vendor's LabVIEW Windows app + manual — reference only, never modify.
- Name & Purpose:
- Quality Gates (MUST pass before handing work back):
- Lint & Format:
.venv/bin/ruff check . && .venv/bin/ruff format --check . - Type / Static Check:
.venv/bin/mypy(strict) - Tests:
.venv/bin/pytest(GUI tests run offscreen; hardware tests are marked@pytest.mark.hardware, excluded by default, run withPS3005D_PORT=/dev/ttyACM0 .venv/bin/pytest -m hardware)
- Lint & Format:
- Domain & Safety Constraints:
- Hardware safety: never exceed 30.00 V / 5.000 A; validate every setpoint in
protocol.pybefore it is sent. The MCP server must additionally enforce user-configurable lower limits (max V, max I) that an agent cannot raise. - Output is off by default: never switch the output on implicitly (on connect, on recall, on setpoint change). Set limits before enabling output. On error, disconnect or shutdown, prefer leaving/turning the output off.
- Single serial owner: only one process may hold the port. GUI and MCP server must not open it simultaneously — share via one process or a local broker, never both opening the port.
- Protocol timing: no terminators; the device garbles replies when paced too fast — keep ≥100 ms after writes and ≥40 ms after queries (measured, see
docs/protocol.md), all port access behind one lock. - Protocol quirks are documented in
docs/protocol.md(e.g.SAV<n>ignores n and writes the active memory). Keepsim.pyfaithful to verified behavior. Tests use the simulator, never real hardware, unless run with-m hardware. - Front-panel memories are user data: never overwrite M1–M5 without the user's consent; hardware tests must restore what they change.
- Must run on Windows and Linux: no POSIX-only APIs; use
serial.tools.list_portsfor port discovery.
- Hardware safety: never exceed 30.00 V / 5.000 A; validate every setpoint in
- Documentation Map (link instead of duplicating content):
README.md— project overview, install, usage (screenshots indocs/images/, regenerate with the simulator).docs/protocol.md— wire protocol, firmware quirks and verification status.docs/mcp.md— MCP tools, safety model, client setup (standalone and GUI-hosted).CONTRIBUTING.md— contributor guidelines;CHANGELOG.md— update with every user-visible change.PC_SOFTWARE/— vendor app + manual, local only (Velleman's copyright, git-ignored).- CI:
.github/workflows/ci.ymlruns the quality gates on Linux + Windows, Python 3.11–3.13.
- Git Hygiene:
- Commit only when explicitly asked by the user.
- Never commit
.agents/memory/, build artifacts, local captures, or temporary debug files.
2. Security & Secrets Policy (CRITICAL — ALWAYS ENFORCED)
- Zero Secrets in Git: NEVER hardcode, stage, or commit secrets, API keys, tokens, passwords, private keys, or credentials into Git.
- Environment & Secret Files: Store all sensitive values in
.envor.secrets/(which must be git-ignored via.gitignore). Only commit sanitized templates with placeholder values (such as.env.example). - Agent Memories & Docs: NEVER write raw secrets, tokens, or credentials into
.agents/memory/*.md, artifacts, logs, or documentation. - Pre-Commit / Review Gate: Always verify
git statusandgit diffcontain no secrets or unintended files before finishing or approving any task.
3. Execution Mode & Multi-Agent Orchestration
- Direct Execution (Small Tasks): For single-file edits, straightforward bug fixes, or quick questions, execute the task directly in the main session while following Sections 1 and 2.
- Team Orchestration (Complex Tasks): For multi-file features, architectural changes, unfamiliar subsystems, or broad audits, act as Lead Orchestrator and load
.agents/skills/team-orchestration/SKILL.md. - Persistent Memory: Local agent state lives in
.agents/memory/(git-ignored). For non-trivial tasks, read.agents/memory/project_state.mdfirst (runbash .agents/init_memory.shif.agents/memory/does not exist yet).
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.

