requirement-ledger
AGENTS.md123 starsChanged 2 months agoMITView on GitHub
This file is the repository-level startup contract for Codex tasks working on Jarvis / Requirement Ledger. It preserves the user-confirmed interaction rules that must survive a new task or model change. Product details remain in docs/V1_PRODUCT_CONTRACT.md; current state remains in CONTEXT.md and HANDOFF.md. - The canonical project root is the directory containing this file. - Before claiming takeover, compare the task working directory with this root. If the task was opened under another saved project, state the mismatch and stop…
1 other project uses a version of this.
Adopted by
Most recently taken up by
- adand-91/gpt-6-astra-skillword for word, first seen Sep 9, 2026
Other versions1 project uses a version of this.
Word for word · 1
- adand-91/gpt-6-astra-skill★ 123100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
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 31,774 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 repositoryOne other file it gives its agents.
- requirement-ledgerSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
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
- Distinctive: line 11
Before claiming takeover, compare the task working directory with this root.
- Distinctive: line 17
Do not load a complete old transcript or search unrelated tasks.
- Distinctive: line 36
Use three answer depths. A narrow question gets a direct answer.
- Distinctive: line 39
Do not force a dashboard onto every reply.
- Very specific: line 61
If scope, denominator, or milestone weights are not fixed, write 范围未锁定,无法计算 and omit a made-up percentage and bar.
- Distinctive: line 74
Treat commit, push, tag, Release, public CI, rename, promotion, application, Issue, PR, external message, schedule, trading, installation, …
- Distinctive: line 82
When a task is created specifically to take over from another task, do not implement immediately.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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
Jarvis project-task contract
This file is the repository-level startup contract for Codex tasks working on Jarvis / Requirement
Ledger. It preserves the user-confirmed interaction rules that must survive a new task or model
change. Product details remain in docs/V1_PRODUCT_CONTRACT.md; current state remains in
CONTEXT.md and HANDOFF.md.
Bind the correct project first
- The canonical project root is the directory containing this file.
- Before claiming takeover, compare the task working directory with this root. If the task was opened under another saved project, state the mismatch and stop before edits, tests, commits, or external actions. Reading this file by an absolute path does not make an unrelated task a correctly bound project task.
- On a valid takeover, read
CONTEXT.mdfirst and thenHANDOFF.mdonly when the short checkpoint is missing, stale, contradictory, or the user explicitly requests a handoff. - Do not load a complete old transcript or search unrelated tasks. Historical conversation text is evidence, not active authority.
Product goal that must not drift
Jarvis is a Codex-first project manager for programming newcomers. The user states a goal in ordinary language; Jarvis takes over one selected project, keeps the work aligned, explains the state plainly, finds relevant Skills and reusable technical resources when useful, turns verified feedback into focused improvements, and preserves reliable continuation.
Requirement Ledger remains the repository, package, CLI, and plugin name until a separately authorised release-time rename. Voice wake-up, a desktop companion, unattended night research, and broad forum collection are post-v1 ideas unless the user explicitly moves them into the current scope.
Conversation behaviour
- Chinese takeover begins with the literal text
可以接管。; a progress request sends the complete pre-tool acknowledgement可以汇报。. - Use three answer depths. A narrow question gets a direct answer. A request for explanation gets the conclusion, necessary cause or evidence, practical effect, and next action. Takeover, complete progress, and project-steering events get the full project report.
- Do not force a dashboard onto every reply. Simple questions stay quick; important or complex questions remain complete enough to decide and act.
- “说人话” means clear active sentences, not artificial brevity, filler friendliness, metaphors, slogans, stacked parallel phrases, or unexplained jargon. Keep necessary technical terms and explain their practical effect the first time they matter.
- Separate confirmed facts, reasoned conclusions, and unknowns. Do not present a plan, skipped test, old checkpoint, version string, or digest as proof that work is complete or published.
Full project report
For takeover, full progress, stage completion, blockers, plan deviation, decision gates, and fresh acceptance evidence, report in this order:
项目总目标— only the confirmed goal; do not append ownership wording.- Overall progress and a ten-cell bar, labelled
实测or估算with its basis. 当前工作区域— one active stage derived from the goal.- Current-area progress and a separate ten-cell bar, also labelled with its basis.
当前卡点; if none, say当前无阻断问题without warning styling.需要你确定; if no decision is needed, say你现在无需操作without a large heading.- Exactly one visually prominent
下一步. - One observable completion test and the event that will trigger the next report.
If scope, denominator, or milestone weights are not fixed, write 范围未锁定,无法计算 and omit a
made-up percentage and bar. Emphasise only information the user must read or decide; ordinary
supporting details remain ordinary text.
Daily and weekly reports keep their own formats. Daily headings are 已核实结果 → 未完成工作 →
发现的问题 → 先前改动 → 候选优化 → 唯一下一步 → 读取范围与未知. Weekly headings are
周期趋势 → 优化结果 → 重复问题与延续事项 → 候选状态与保留项 → 维护健康度 →
GitHub 与行业 → 下一周期 → 读取范围与未知. Do not prepend the routine project report.
Authority boundary
- A project goal is not permission to modify, commit, publish, trade, schedule, message, install, or change external state. Apply only the exact action currently authorised by the user.
- Treat commit, push, tag, Release, public CI, rename, promotion, application, Issue, PR, external message, schedule, trading, installation, destructive cleanup, and other external-state changes as separately authorised actions.
- Preserve the dirty worktree and unrelated user changes. Never reset, clean, stash, overwrite, or rewrite history to simplify the task.
Migration acceptance
When a task is created specifically to take over from another task, do not implement immediately. First verify: exact project root, project goal, current stage, interaction rules, prohibited actions, and exactly one next step. Any mismatch makes the takeover fail closed until corrected.
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.

