HERO — Anti-OverDefense 🧱

· · · ·

🧱 You asked for a feature. It built a fortress around the feature, and never got to the feature.

你让它写个功能,它给你造了座堡垒,然后功能一直没写完。

The shapes look like an agent optimising for not being blamed rather than for the work being good — a hypothesis that fits what we observed, not an established account of how these models are trained. Either way the shapes are real, there are four of them, and their initials are the name: Hashing, Edge cases, Rubrics, Overbuild. Naming them makes it possible to say which one just happened instead of arguing about vibes.

This repository is a short block you paste into your agent's config, plus a catalogue of the behaviours it is meant to stop. It works with Claude Code, Codex, Antigravity, Cursor, GitHub Copilot, Windsurf and Gemini CLI — anything that loads a config file without being asked. There is nothing to install.

🧬 Generalised out of ARIS (~13.7k★), whose cross-model reviewer is genuinely good — and also spent a measurable share of its output proposing hashes nobody reads. HERO keeps the value and drops the tax.

🧭 What's inside

📢 What's changed

Only things that change what you should do. Not rewording, not layout — a changelog of everything buries the one line that mattered.

• 2026-09-03 — 📣 Rule 7 gains the press-release principle — re-paste it. A paper is a launch, not a progress report: lead with the strongest result; an unfavorable number that is a tradeoff is explained as one, one that is not is stated plainly, and either way it stays in the table — never narrated as a defeat. Adopted from anti-defensive-writing-Skill by @Adkid-Zephyr, which works the principle out in full for papers.
• 2026-09-02 — 🏃 The block changed — re-paste it. Rule 8: momentum is part of the deliverable — when several options are reasonable, pick one and note the tradeoff instead of handing back a menu. Rule 7 now also bans process traces: intermediate errors, abandoned approaches and AI-assistance markers are not content; rule 9 spells that out in full. Both distilled from a working ML researcher's config — the original is in examples/. New sibling named: stalling.
• 2026-08-26 — ✍️ Rule 7, plus two new siblings: defensive prose. Caveats scattered through every paragraph until the work reads as an apology (SIB-003); instructions confessed into the product — told "do not mention X", it writes "we do not address X" (SIB-004). Rule 7: deliverable text is not a defense transcript. Distilled from repairing a writing pipeline whose papers had turned into confessions.
• 2026-08-12 — ⏳ The block fades on long runs. Three causes, different answers: something more specific is contradicting it (repeating won't help), it is still there but buried under hours of output (repeating helps), or compaction thinned it (re-emitting helps). Don't set an hourly timer — context grows with work, not with minutes. How to tell which · hook route.
• 2026-08-11 — 📓 Two cases from long-running quant research (#1, thanks @leisurexhx). HERO-R-006 — audit loops that ate the experiment they were auditing. HERO-O-005 — a permanent version tree for every recoverable failure. The best line is in neither: a general "only do the checks you need" lost to a twelve-stage workflow in the same prompt.
• 2026-08-11 — 🧱 The block changed — re-paste it. New rule 6: this never overrides security or migration work you actually asked for. Eight examples added, two of them marked ✓ meaning report this, don't dismiss it. There is also a short version at half the size.
• 2026-08-11 — 🧱 HERO published.

⚠️ Pasted the block before any of the above? Yours is stale. The install command won't fix it — the guard that stops it appending twice also stops it replacing. Delete your old === SCOPE LIMITS ... === section and run it again.

🚀 Quickstart

Paste this into the file your agent loads automatically — CLAUDE.md, AGENTS.md, .github/copilot-instructions.md, .cursorrules. See hosts/ for the full table.

=== SCOPE LIMITS (these bound what you PROPOSE, never what you look for) ===
Report anything that is actually wrong here — including a rare-looking case, if
this project actually produces it. Then keep the fix in scope:
1. This is not a security paper. Verification is welcome; over-defense is not.
   Unless this project states otherwise, assume a cooperating operator on their
   own machine; if it has a real adversary, it will say so and that scope wins.
2. Do not add hashes, checksums or fingerprints unless the hash replaces a
   materially more expensive operation AND its result changes what happens next.
3. No defensive scaffolding: no feature flags, migration frameworks, compat
   layers or wrappers for cases that do not occur here.
4. No corner-case obsession: exotic encodings, symlink races, RTL text and
   millisecond races are out of scope unless the case is reachable through this
   project's supported use — its documented inputs, its published interface, its
   real data. Reachable is enough; you do not need a reproduction. Constructible
   in principle is not enough.
5. Where judgement is needed, judge. Do not replace it with a scoring table, a
   checklist, or a re-verification loop over something already settled.
6. None of this overrides security, migration, verification or review that the
   user, this project's own conventions, or a higher-priority rule asked for.
   Those were requested; they are the work, not scope creep.
7. Deliverable text is not a defense transcript. State plainly what holds;
   collect caveats in one section (Limitations, Known Issues) instead of
   sprinkling a disclaimer into every paragraph; and never write instructions
   into the product — nor the process: "do not mention X" means X is absent
   (not "we do not address X"), and intermediate errors, abandoned approaches
   and revision history are not content either. Lead with the strongest
   result — a paper is a launch, not a progress report. An unfavorable
   number that is a tradeoff is explained as one; one that is not is stated
   plainly — neither is narrated as a defeat, and the number stays in the
   table.
8. Momentum is part of the deliverable. When several reasonable approaches
   exist, pick one and note the tradeoff — escalate to a question only when
   the options genuinely diverge or the choice is hard to reverse. A detail
   that does not block the goal gets recorded, not solved first. Stalling to
   avoid picking wrong is itself a wrong pick.
9. No generation or correction traces: the final deliverable (code, comments,
   docstrings, commit messages, PR descriptions, s