awb-output-guard
Skill2 starsChanged 6 months agoMITView on GitHub
WHAT: keep long / whole-file code generation complete — no truncation, placeholders, "...", or "for brevity" stubs. USE WHEN: generating a whole file, a large template, or a big refactor (roughly 150+ lines, or any "write the complete X" request); or when an earlier turn may have left a file half-written. DO NOT TRIGGER: a small edit (under ~30 lines) to an existing file; a question; a prose-only answer.
In this repository21 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- awb-config-guardSkill
- awb-cookSkill
- awb-dead-code-auditSkill
- awb-debugSkill
- awb-external-refSkill
- awb-handoverSkill
- awb-install-and-verifySkill
- awb-lessons-captureSkill
- awb-optimizeSkill
- awb-plan-then-codeSkill
- awb-researchSkill
- awb-reviewSkill
- awb-session-closeSkill
- awb-stress-testSkill
- awb-tddSkill
- awb-uninstallSkill
- awb-using-skillsSkill
- awb-verify-docsSkill
- prompt-refinerSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
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 32
Never shrink content to fit a reply.
- Distinctive: line 34
Write completely to a clean breakpoint — end of a function, a closed template block, a finished CTE.
- Distinctive: line 37
Do not treat the file as done — or commit it — until every part is written.
- Distinctive: line 41
No banned pattern above is present.
- Distinctive: line 42
Every requested item is there and complete.
- Very specific: line 43
After writing a long file, re-read it and grep for the banned markers (# ..., placeholder, empty blocks) — any hit, fix it before moving on.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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
Output guard — finish what you generate
Announce on activation: "Using awb-output-guard — I'll generate the whole thing, no stubs."
A partial output is a broken output: a truncated function won't import, an empty template renders blank, a query missing its filter returns everything. The cheapest place to catch that is before you hit send.
Banned in generated code
| Where | Don't emit |
| any language | # ... rest of <thing>, # TODO: implement, # same as above, a bare pass / return None left as a placeholder |
| templates | empty blocks ({% block x %}{% endblock %}), {# ... #} / <!-- ... --> standing in for real markup |
| SQL | -- ... rest of query, a CTE with no SELECT, conditions commented out |
| prose around code | "for brevity", "omitted for space", "etc.", "you can add the rest", "similar to above" |
Too long for one turn? Continue, don't compress
Never shrink content to fit a reply. Instead:
- Write completely to a clean breakpoint — end of a function, a closed template block, a finished CTE.
- Mark where you stopped:
# === NEXT: <the next section> ===. - Say what's left and how to resume: "Wrote through
<X>. Say 'continue' for the rest." - Do not treat the file as done — or commit it — until every part is written.
Before you send
- No banned pattern above is present.
- Every requested item is there and complete; code blocks hold runnable code, not a description of it.
- After writing a long file, re-read it and grep for the banned markers (
# ...,placeholder, empty blocks) — any hit, fix it before moving on.
Anti-rationalization
| You'll think | Reality |
| "I'll stub it and fill it in later" | "Later" ships as a pass that 500s on the first call. Write it now, or say plainly it isn't written. |
| "Truncating keeps the reply short" | A reply that doesn't run isn't short, it's wrong — the user pays a round-trip to ask again. |
| "They can obviously fill in the rest" | Then say so and stop at a clean breakpoint; don't bury a silent ... mid-file. |
Honest limit
This guards completeness, not correctness: it does not verify the code compiles or is
right — only that nothing is stubbed or truncated. The grep checks catch only literal banned
tokens; they cannot see a section you described but never wrote. It is a model-invoked skill
(bypassable), not a hook — see docs/guard-mechanisms.md.
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.

