Progress

/progress - Knowledge Progression Tracking

Track how your understanding of a topic evolves across sessions. A progression is a sequence of documents (baseline, deepenings, corrections, pivots) that captures the full arc of investigation.

Usage

/progress [args]

Subcommands

start - Begin a new progression

Start tracking a new topic.

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-init --project PROJECT --topic "Topic Name"

Determine PROJECT from the current working directory (basename of CWD). Ask the user for the topic name if not provided.

Example: /progress start ECS Task Placement Strategy

To create a progression for a different project (cross-project), pass --project:

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-init --project cloudfix --topic "Topic Name"

Use _global for progressions not tied to any project:

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-init --project _global --topic "AWS Cost Patterns"

add - Add a document to a progression

Save the current analysis/finding as a numbered document in the progression.

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-add \
  --project PROJECT \
  --topic "TOPIC" \
  --number NN \
  --title "Document Title" \
  --type TYPE \
  [--file PATH] \
  [--corrects NN]

Document types:

• baseline — Initial understanding, first pass
• deepening — Deeper analysis that builds on previous docs
• correction — Corrects a previous document (use --corrects NN)
• pivot — Fundamental change in direction or approach
• synthesis — Consolidation of multiple findings

How to determine the number: Look at the existing progression with pi-progression-status and use the next sequential number.

How to create content: Pipe content directly via stdin using --file -. Do NOT write to /tmp. Example:

cat <<'DOC' | pi-progression-add --project PROJECT --topic TOPIC --number NN --title "Title" --type TYPE --file -
# Document Title

Content goes here...
DOC

The content should capture:

• What was discovered/analyzed
• Key data points or evidence
• How this relates to previous documents in the progression

Example: /progress add correction "Actual Cost is $3.9K not $387K" --corrects 1

correct - Mark an existing document as corrected

Use this when a new finding invalidates a previous document. This is a shortcut that combines add with --type correction --corrects NN.

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-add \
  --project PROJECT \
  --topic "TOPIC" \
  --number NN \
  --title "Correction Title" \
  --type correction \
  --corrects PREV_NN \
  --file /path/to/content.md

Example: /progress correct 1 "CUR shows actual cost is much lower"

conclude - Mark a progression as complete

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-conclude --project PROJECT --topic "TOPIC"

Concluded progressions are no longer injected into session context but remain searchable.

Example: /progress conclude ECS Task Placement Strategy

show - Show details of a specific progression

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-status --project PROJECT --topic "TOPIC"

Shows: status, document list with types, corrections, current position.

Example: /progress show ECS Task Placement Strategy

list - List all progressions for the project

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-status --project PROJECT

Shows all progressions with their status (active/concluded/parked).

Add --all to list progressions across all projects:

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-status --all

Example: /progress list

search - Search progressions across all projects

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-search QUERY [--project PROJECT] [--limit N]

Searches all progression documents via FTS5 full-text search. Without --project, searches globally across all projects.

Example: /progress search "cost optimization" Example: /progress search "migration" --project cloudfix

To reindex existing progressions (run once after upgrading):

${CLAUDE_PLUGIN_ROOT:-~/.claude/project-intelligence}/bin/pi-progression-search --reindex

Guidelines

• One progression per investigation arc. A new topic or completely separate question gets its own progression.
• Number documents sequentially starting from 00. Use pi-progression-status to find the next number.
• Always mark corrections explicitly. When new data contradicts a previous document, use --type correction --corrects NN so the progression tracks what was wrong and why.
• Write content that captures reasoning, not just conclusions. Future sessions need to understand why you reached a conclusion.

Active Progressions in Context

Active progressions are automatically injected into session context. They include:

• Topic name and document count
• Current position summary
• Corrections (what was wrong)
• Open questions (what to investigate next)

This helps new sessions pick up where previous ones left off without re-reading all documents.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: theaichimera
• Source: theaichimera/claude-code-project-intelligence
• License: MIT

Install and usage instructions live in the source repository linked above.