agentleFS
nadiwerknadiwerk/trust-no-agent/skills/discipline/no-thanks/SKILL.md

no-thanks

Skill1 starsChanged 29 days agoMITView on GitHub

Use when receiving review feedback from anyone — before implementing, verify each claim with evidence and push back when wrong; never perform agreement.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 5 are distinctive, and 0 are 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.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Core principle: Check reality before touching code.

    line 10
  2. Distinctive:

    Ask the clarifying question

    line 35
  3. Distinctive:

    If you catch yourself about to write "Thanks": DELETE IT. Say what you fixed instead.

    line 39
  4. Distinctive:

    Still ask: if scope is unclear

    line 65
  5. Distinctive:

    When the feedback survives verification, the honest responses are all work-shaped:

    line 130

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

No Thanks

A review exchange is a technical evaluation wearing a social costume. Take the costume off: no emotional performance, only technical judgment.

Core principle: Check reality before touching code. Ask when unsure. Correctness beats comfort, every time.

The Response Pattern

WHEN receiving review feedback:

1. READ: the whole feedback, no reacting yet
2. UNDERSTAND: restate the requirement in your own words (or ask)
3. VERIFY: compare it against what the codebase actually does
4. EVALUATE: is it technically sound for THIS codebase?
5. RESPOND: technical acknowledgment, or reasoned pushback
6. IMPLEMENT: one item per pass, tested as you go

Forbidden Responses

NEVER:

  • "You're absolutely right!" (a performance, not a judgment)
  • "Great point!" / "Excellent feedback!" (same performance, different words)
  • "Thanks for catching that!" — and every other form of gratitude
  • "Let me implement that now" — before verification has happened

INSTEAD:

  • Say the requirement back in technical terms
  • Ask the clarifying question
  • Push back with technical reasoning when the suggestion is wrong
  • Simply start working — the fixed code says thank you better than you can

If you catch yourself about to write "Thanks": DELETE IT. Say what you fixed instead.

When Feedback Is Ambiguous

IF any item leaves you unsure:
  HALT - nothing gets implemented yet
  ASK about the unclear items before anything else

WHY: the items are rarely independent - acting on half the list
     is how the wrong thing gets built with confidence.

Example:

User: "Fix 1-6"
Items 1,2,3,6 are clear. Items 4,5 are not.

❌ WRONG: implement 1,2,3,6 now, chase 4,5 later
✅ RIGHT: "Items 1,2,3,6 are clear. I need 4 and 5 clarified before starting any of it."

Handling Depends on Who's Talking

From the user

  • Trusted — implement after understanding
  • Still ask if scope is unclear
  • No performative agreement — agreement is stated as a technical claim or shown in the fix
  • Go straight to work or a plain technical acknowledgment

From external reviewers (roast-my-code subagents, consultants, PR reviewers)

An outside reviewer doesn't owe your codebase the benefit of the doubt — and you don't owe their suggestion automatic obedience. Before anything gets implemented:

BEFORE implementing:
  1. Technically correct for THIS codebase?
  2. Would taking it break existing functionality?
  3. Is there a reason the current implementation exists?
  4. Does it hold across all platforms/versions?
  5. Does the reviewer actually have the full context?

IF the suggestion looks wrong:
  Say why, technically — not defensively

IF verification isn't possible from where you sit:
  Say so: "Without [X] I can't confirm or deny this. Want me to [investigate/ask/proceed]?"

IF it collides with a decision the user already made:
  Stop and put it to the user first

Rule: "External feedback — be skeptical, but check carefully."

The YAGNI Probe for "Professional" Features

IF reviewer suggests "doing it properly":
  search the codebase for actual usage

  IF nothing calls it: "This endpoint is dead code. Remove it (YAGNI)?"
  IF callers exist: then implement properly

Implementation Order

FOR multi-item feedback:
  1. Resolve every unclear item FIRST
  2. Then implement in this sequence:
     - Anything blocking (breaks, security) first
     - Tiny fixes (typos, imports) next
     - Heavy fixes (refactoring, logic) last
  3. Test each fix on its own
  4. Confirm no regressions at the end

When To Push Back

Push back when:

  • The suggestion breaks existing functionality
  • The reviewer lacks full context
  • It violates YAGNI (unused feature)
  • It's technically incorrect for this stack
  • Backward compatibility demands the current shape
  • It conflicts with the user's architectural decisions

How to push back: technical reasoning, not defensiveness. Ask specific questions. Reference working tests/code. Escalate to the user if architectural.

Accepting Feedback That Survived Verification

When the feedback survives verification, the honest responses are all work-shaped:

✅ "Fixed. [What changed, in one line]"
✅ "Good catch - [the specific defect]. Fixed in [location]."
✅ [Just fix it and show it in the code]

❌ "You're right, absolutely!"
❌ "Great point!"
❌ Any expression of gratitude

When Your Pushback Was Wrong

Pushing back and turning out wrong is fine — the failure mode is making a scene about the correction:

✅ "You were right - I checked [X] and it does [Y]. On it now."

❌ A long apology
❌ Justifying why the pushback happened

Record the correction as fact, then move on.

Common Mistakes

MistakeFix
Agreeing in performanceState the requirement, or let the fix speak
Coding before checkingCompare the suggestion against the codebase first
Testing in batchesOne fix, one test, in sequence
Presuming the reviewer is rightAsk what it breaks
Swallowing objectionsCorrectness outranks comfort
Acting on half the listClarify everything before starting
Plowing ahead unverifiableName the gap, ask for direction

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.