| name | progress |
| description | Track evolving understanding of a topic through a sequence of documents |
| user_invocable | true |
/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 <subcommand> [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 -
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.