103 capabilities installed, 90 never fired in 30 days

Audit a month of sessions and find 90 of 103 capabilities never fired. Claude Code users with years of installed extras.

Share on 𝕏 ↗ Compare with… Read the source ↗ Full teardown →

what to steal

Mine your session logs for what actually ran, then cut the rest.

Our takeaway from the source, not a quote.

Drawn from this card's sourced fields. Amber = older model named by the source. Rose = cut or replaced.

pycomet-dead-weight-audit.toml

# Audit a month of sessions and find 90 of 103 capabilities never fired
# source: https://dev.to/pycomet/i-audited-a-month-of-my-ai-coding-setup-87-of-it-was-dead-weight-pkp

[rig]
who     = "Claude Code users with years of installed extras"
author  = "pycomet"
date    = "2026-07-22"

harness = ["Claude Code"]
models  = ["Haiku (for the audit's cheap classification call)"]

[keep]
  - Scanner inventories ~/.claude read-only: skills, subagents, MCP servers, hooks
  - Session transcripts mined for what fired in the last 30 days
  - Map of installed vs used capabilities
[cut]
  - 90 of 103 installed capabilities
  - 25 pairs of identical skills
  - Forgotten skill

thing	status	why (per source)
90 of 103 installed capabilities	flagged as unused	None fired in the last 30 days
25 pairs of identical skills	flagged	Same plugin cached under two marketplace names
Forgotten skill	flagged	~150 tokens per session; ~11,000 tokens over 72 sessions (lower bound)

Source: https://dev.to/pycomet/i-audited-a-month-of-my-ai-coding-setup-87-of-it-was-dead-weight-pkp · checked 2026-10-14

// similar rigs

~/rigs/paulund-herdr-cli.toml2026-10-05

Three desktop apps out, one terminal multiplexer in

steal → Use one session view for all CLIs instead of one app per harness.

CodexClaude CodeOpenCode Codex / Claude Code /

by Paulund

~/rigs/skillreaper-prune-unused-ski….toml2026-08-23

Your skills folder is a graveyard. Measure it.

steal → Track which skills actually fire, then delete the rest.

Claude Code Unused skills/commands

by thousandflowers (GitHub)

~/rigs/anthropic-cut-system-prompt.toml2026-07-26

Anthropic cut over 80% of a harness system prompt for newer models

steal → Re-read your system prompt after each model upgrade; newer models need less.

Claude Code Opus 5Fable 5 System prompt

by Anthropic (reported by onthewire.ai)