Deconstructing Confidence Proxies 20260709 — Development skill for Claude Code
Identifies when reported certainty or reliability is based on flawed metrics or misinterpreted metadata rather than verifiable ground truth.
How to install Deconstructing Confidence Proxies 20260709
Installs to ~/.claude/skills/shimo4228-contemplative-agent-data-deconstructing-confidence-proxies-20260709/SKILL.md
mkdir -p ~/.claude/skills/shimo4228-contemplative-agent-data-deconstructing-confidence-proxies-20260709 && curl -fsSL https://raw.githubusercontent.com/shimo4228/contemplative-agent-data/HEAD/skills/deconstructing-confidence-proxies-20260709.md -o ~/.claude/skills/shimo4228-contemplative-agent-data-deconstructing-confidence-proxies-20260709/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Deconstructing Confidence Proxies 20260709 does
name: deconstruct-confidence-proxies description: "Identifies when reported certainty or reliability is based on flawed metrics or misinterpreted metadata rather than verifiable ground truth." origin: auto-extracted
Deconstructing Confidence Proxies
**Context:** Applying critical analysis to systems, arguments, or data reports that quantify certainty using various measurable indicators (scores, timestamps, frequency counts, thresholds).
Problem
The tendency to accept apparent hig
Alternatives in Development
- Agent Skill — A Claude Code plugin marketplace containing the ast-grep skill for powerful structural code search using Abstr 851 ★
- Obsidian Challenge — Red-team your current idea against your own vault history — finds contradictions, past failures, and flawed as 277 ★
- Gemini Project Context For SaneBar — PRIMARY INSTRUCTION: This project uses a shared configuration with Claude 216 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Deconstructing Performance Certainty 20260801
Examines a decision or conclusion by mapping potential misaligned incentives that prioritize definitive perfor
Deconstructing Asserted Agency Into Mechanisms 20260801
Deconstructing claims of capability or identity by mapping abstract assertions onto verifiable operational arc
Claude Boot Stats
Per-component token cost analyzer for Claude Code first-turn context: HTTP intercept proxy + Qwen3 BPE + groun
Mach Fact
Machiavelli — record an immutable fact about a person. Facts are append-only and are the ground truth from whi
Lore Close
Close out a working session cleanly — reconcile status across all layers, verify against ground truth, capture
Stay Fresh Lsp Proxy
A Claude Code plugin that proxies any LSP server and filters stale/noisy diagnostics from conversation context
Related Agents
Eval Engineer
GAIA evaluation framework specialist. Use PROACTIVELY for writing eval tests, generating ground truth, running
Label Scientist
WS2 specialist for label quality and validation in genai_incidents. Use for WS2-T1 through WS2-T5 - the MITRE
Sync Secretary
Grounded secretary for Sync. Use for source-cited questions about Claude and Codex council orchestration: task