Transcript miner
Portable Agent Skills for non-RAPP systems and launchpads for single-file RAPP agents
npx -y skills add kody-w/rapp-skills --skill transcript-minerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 22 days oldThe repository was created 22 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Mine Claude Code session history for usage patterns, mistakes, and automation candidates. Use when Kody says "audit my sessions", "what am I doing wrong in Claude Code", "usage audit", "mine my transcripts", "analyze my Claude Code history", "what should be a skill", or asks how he's using Claude Code across past sessions. Extracts tool stats, error signatures, Read:Edit ratios, my-message categories (corrections/rejections), and permission denials from ~/.claude/projects JSONL — with evidence, never vibes.
SKILL.md
8.3 KB, ~4.8k tokens by cl100k_base, as published. Nobody here has run it
transcript-miner
Turns the raw ~/.claude/projects/<project>/<session>.jsonl archive into a
usage audit. One JSON message per line (type, content blocks, tool_use,
timestamps). Sessions are large — this NEVER reads whole transcripts into
context; it runs a streaming extractor and returns aggregates.
Run it
python3 ~/.claude/skills/transcript-miner/scripts/mine.py --window 40 --json /tmp/audit.json
Flags: --window N (last N substantial sessions), --min-bytes N (size floor,
default 200KB), --project SUBSTR (filter to one project), --json OUT (full
per-session rows).
Turn signals into findings
The script gives you the numbers; you write the audit. Rules:
- Every claim cites a session file + excerpt. No uncited findings.
- Rank by frequency × cost, label single-anecdote vs recurring.
- Findings schema:
{finding, evidence, frequency, impact, confidence, fix}. - Fixes are behavioral rules, not principles — "never
cd, pass absolute paths" not "be tidy". Each recurring failure → one rule / skill / hook. - Read:Edit ratio: >6 = research-first (good), <2 = edit-first — but bulk file-generation sessions (estate sweeps) skew it low legitimately; note that.
- Always include a "could not verify / out of scope" section.
Output
An evidence-backed report + draft SKILL.md for the top skill candidates + hook configs for the top automations. See the P2 audit for the reference shape.
<!-- toaster:generated:begin -->Deterministic steps
Lifted verbatim from the procedure above by toaster.py toast. Run them in order, substituting the typed parameters; do not paraphrase:
cd
python3 ~/.claude/skills/transcript-miner/scripts/mine.py --window 40 --json /tmp/audit.json
What ships with it: 2 files
17.5 KB alongside SKILL.md, 2 of them executable
scripts/
- mine.pyruns5.2 KB
- transcript_miner_agent.pyruns12.3 KB
Gives 0 of the 12 instructions most video audio skills give in ~4.8k tokens
Counted across 622 of the 795 authors here whose files we hold, read 2026-08-07
- Read individual rule files for detailed explanationsin 21 of 622, across 10 files
- Render final videoin 13 of 622, across 6 files
- Use WAV PCM 16kHz mono audio formatin 12 of 622, across 3 files
- Use this skill when dealing with Remotion codein 11 of 622, across 4 files
- Save generated audio to a WAV filein 11 of 622, across 4 files
- Handle conversion errors gracefullyin 10 of 622, across 6 files
- Add captions to videos alwaysin 10 of 622, across 4 files
- Generate music from text descriptions using MusicGenin 9 of 622, across 2 files
- Do not skip pipeline layersin 9 of 622, across 3 files
- Do not make one tool do everythingin 9 of 622, across 3 files
- Use Azure Document Intelligence for complex PDFsin 9 of 622, across 4 files
- Never ask the user to paste their full API keyin 9 of 622, across 3 files
Said here and by no other author read
- run the streaming extractor script
- cite a session file and excerpt for every claim
- rank findings by frequency and cost
- label single-anecdote versus recurring findings
- write fixes as behavioral rules not principles
- note bulk file generation skewing read-to-edit ratio
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.