SKILLEMALL.ai
GitHub ↗ RU Sign in

AC confirmation-bias

Activate when: user says 'we keep finding evidence that supports our view,' 'the team is all aligned on this,' 'I've done the research and it checks out,' or a decision moves forward with only supporting evidence cited. Do NOT activate when: context is explicit advocacy (legal brief, pitch deck) where one-sided argument is the design; or stakes are too low to justify structured disconfirmation. More: deciqai.com/c/confirmation-bias

ClawHub Agent Skills author: deciqAI v1.0.4 MIT-0 6 files body ≈ 1 644 tokens Open the sourceclawhub.ai↗ analyzed 2 d ago

Activate when: user says 'we keep finding evidence that supports our view,' 'the team is all aligned on this,' 'I've done the research and it checks out,' or…

As a process C 63/100 · Has gaps — weak spots: inputs and preconditions, failures and branches, running it twice

ProcedureFinanceAI and agentsResearchtype and topics are labelled automatically from the skill text
Runs in: OpenClaw Claude / Claude Code ChatGPT / Codex Gemini CLI Cursor GitHub Copilot / VS Code Qwen Code Kimi Code CLI Z.ai (GLM) DeepSeek (Deep Code) Mistral Vibe LangChain (deepagents) Goose OpenCode
JSON
Technical rating
A
95/100
safety, quality, tests
Safety 60%
95
Quality 40%
95
Run on models
none yet
Process rating
C
63/100
Has gaps
Inputs and preconditions w 11
0
Failures and branches w 10
0
Progress reporting w 2
0
the three weakest of ten parameters · all ten

How to improve

    For the model run — optional
    • Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
    • A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.

    Guard findings · 5

    ✓ No critical or high findings

    Medium and low: 5
    • low Risky intent intent-offensive-security examples/ai-thesis-confirmation-2023-2026.md:17
      Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (test fixture / example file; quoted — discussed, not commanded)
      **Step 5 — Install structural countermeasure.** Personal vigilance ("we're being rigorous") does not survive contact with a motivating narrative, so the fix is structural. Concretely: a rotated, manda
      fixturequoted
    • low Risky intent intent-offensive-security examples/ai-thesis-confirmation-2023-2026.md:26
      Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (test fixture / example file)
      5. Structural countermeasure: rotated red team, pre-mortem, blind eval, falsification-first (three refuting cases before one confirming demo), eval owned by a disinterested party
      fixture
    • low Risky intent intent-offensive-security examples/peter-wasons-2-4-6-task-1960.md:45
      Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (test fixture / example file; quoted — discussed, not commanded)
      This last finding has the most operationally important implication: **individual debiasing efforts are weakly effective; structural intervention through adversarial process is strongly effective.** He
      fixturequoted
    • low Risky intent intent-offensive-security SKILL.md:39
      Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (quoted — discussed, not commanded)
      5. Close: name the falsification test + structural countermeasure (Devil's advocate, red team, blind evaluation).
      quoted
    • low Risky intent intent-offensive-security SKILL.md:52
      Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (quoted — discussed, not commanded)
      **Step 5 — Install structural countermeasure:** Devil's advocate (rotated, mandatory) · Red team · Pre-mortem (Klein 2007) · Blind evaluation · Falsification-first design (three refuting cases before
      quoted

    Files scanned: 6. Evidence is masked. Grey chips explain why severity was lowered.

    Against the Agent Skills spec

    ✓ No remarks against the Agent Skills spec

    Process rating: all ten parameters 63/100

    • 0Inputs and preconditions. Does not say what the process needs to start
    • 0Failures and branches. Linear process with no failure handling
    • 0Progress reporting. Says nothing while it works
    • 30Running it twice. 3 mutating operations with no state check
    • 50When it triggers. No condition that starts the skill
    • 60Result and completion. Output format stated, no completion criterion
    • 100Tools and files. No external tools needed
    • 100Steps. 20 steps
    • 100Consistency. Name and required fields are in place
    • 100Execution cost. Instruction body is 1644 tokens

    Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.

    Quality signals

    • +5Description has no quoted example phrases that should trigger the skill
    • +1No license
    • +2Single-language instructions
    • +4Description says when NOT to use the skill
    • +3Description length 435: enough signal without eating the budget
    • +4Structure: 10 headings
    • +3Step-by-step instructions: 20 items
    • +3Output format is stated explicitly
    • +4Has examples (1 code blocks)
    • +4Reference files are cited in the instructions (1 of 1)

    Quality base 70; lint remarks subtract, signals add up to 100. Result: 95.

    External checks

    ClawHub: clean
    This is a markdown-only critical-thinking skill that coaches users to test claims against disconfirming evidence, with no executable behavior or privileged access.
    LLM: benign (high) · VirusTotal: · 16 Jul 2026