Analyzes denied Plannotator plans to extract patterns and build corrective prompts.
npx skills add https://github.com/luan/agents --skill plannotator-compound
Please help me install this Agent Skill. Skill: plannotator-compound Source: https://github.com/luan/agents/tree/main/skills/plannotator-compound Command: npx skills add https://github.com/luan/agents --skill plannotator-compound
It analyzes a user's Plannotator plan archive (or falls back to Claude Code ExitPlanMode logs) to extract denial patterns and convert them into actionable, personalized prompt instructions, then packages everything into a polished HTML dashboard report.
~/.plannotator/plans/ and builds a denial taxonomy, recurring phrases, and an evolution-over-time view.scripts/.EnterPlanMode hook file so future planning sessions automatically incorporate the user's denial patterns.Ask the AI to run the plannotator-compound skill to generate your compound planning denial dashboard report from your local Plannotator archive.
High-intent search queries and answers about installing and using this skill.
Analyzing Claude Code ExitPlanMode denial logs extracts human-authored denial reasons from JSONL files to build a feedback taxonomy and generate a corrective prompt. The parser reads the local log directories to identify recurring rejection patterns and compute actionable planning instructions.
Extracting denial patterns from a plan archive requires exhaustive reading and map-reduce batching across denied items. This process reduces structured denial reasons into ranked taxonomies and evolution insights, rendering the results into a self-contained versioned HTML dashboard.
Corrective prompt instructions can be automatically applied to future planning sessions by writing an improvement hook file. This hook integrates the computed denial taxonomy directly into the EnterPlanMode workflow, ensuring subsequent plans avoid past rejection reasons.
The denial analysis workflow supports incremental HTML reporting based on prior report cutoffs. When a versioned HTML dashboard already exists in the plans directory, the system can generate an updated report that only processes new denied items since the last analysis run.
If Plannotator plan data is missing, the denial analysis falls back to parsing Claude Code JSONL logs. The bundled parser in the scripts directory extracts structured denial reasons directly from the ExitPlanMode logs to ensure the taxonomy extraction and reduction process continues.
Map-reduce batching for denied plan extraction should not be used when the plan archive contains only a single denied item or lacks structured denial reasons. The batched extraction and analytical reduction process requires multi-file archival workflows with strict integrity checks to function correctly.
vessuxAnalyze Plannotator plan archives to extract denial patterns and generate HTML dashboard reports.
RayzerrekAnalyzes denied planning archives and generates an HTML dashboard report with corrective prompts.
mazicimertGenerate on-device usage insights from local activity events with SwiftData.
haabeCompare CLAUDE.md instruction variants against baselines using evaluation reports.