By name: shan description: Dump a Claude Code session transcript as navigable Markdown, inspect CC task lists, or manage skill outfit and library (on/off/move/doctor). Use when user wants to review/analyze/export a session transcript, dump/open task files, or manage skills. allowed-tools: Bash(bun:*)
Shan
Claude Code tooling CLI (named after Claude Shannon).
Execute
Run the shan CLI with provided arguments:
bun x @jasonkuhrt/shan $ARGUMENTS
Commands
transcript print [target]
Render a transcript as a readable conversation log — user messages, Claude's prose, tool calls as one-liners.
Arguments:
| Argument | Description |
|---|---|
[target] |
Session ID/prefix, file path, or omit for interactive picker. |
--show |
Add detail layers: results, diffs, thinking, trace, all (CSV list). |
Output: .claude/transcripts/<session-id>.print.md
Examples:
/shan transcript print # pick a session interactively
/shan transcript print dc8ffe42 # specific session
/shan transcript print --show diffs,results # with edit diffs and tool results
/shan transcript print --show all # everything
transcript dump [session-id]
Convert JSONL transcripts to navigable Markdown with columnar headings for editor outline navigation.
Arguments:
| Argument | Description |
|---|---|
[session-id] |
Optional. Partial UUID prefix (e.g., dc8ffe42). Omit for interactive session selection. |
Output: .claude/transcripts/<session-id>.transcript.md in project directory
Examples:
/shan transcript dump # pick a session and dump it
/shan transcript dump dc8ffe42 # dump specific session
Output Format:
Headings use monospace columns with underscore padding for editor outline alignment:
# `001` `01/25/2026 22:45` `file-history-snapshot` `snapshot______________________`
# `002` `01/25/2026 22:45` `progress_____________` `SessionStart:clear____________`
# `003` `01/25/2026 22:46` `user_________________` `sync__________________________`
transcript analyze [target]
Visualize context window consumption and request patterns. Produces two stacked charts (elapsed time + tokens), dimension tracks (type, skill, tool, cache, model, files, trunc, error, alert, top), a top consumers table, summary, and legend.
Arguments:
| Argument | Description |
|---|---|
[target] |
Session ID/prefix, file path, or omit for interactive picker. |
Output: .claude/transcripts/<session-id>.viz.txt
Examples:
/shan transcript analyze # current session
/shan transcript analyze dc8ffe42 # specific session
task dump [target]
Copy task JSON files from ~/.claude/tasks/ into the project's .claude/tasks/ directory. Provides zero-cost task inspection without consuming CC context.
shan task dump test-schema # Dump entire named list
shan task dump 21b0 # Dump UUID-prefixed session list
shan task dump test-schema@3 # Dump single task
shan task dump "@Scaffold" # Find task by subject substring
shan task dump --md test-schema # Convert to Markdown instead of JSON
shan task dump --all # Include all projects' task lists
shan task dump # Interactive picker (TTY required)
task open [target]
Open task lists or individual tasks in $EDITOR (defaults to Zed).
shan task open test-schema # Open list directory
shan task open test-schema@3 # Open single task file
shan task open --all # Include all projects' lists in picker
shan task open # Interactive picker (TTY required)
Task Target Syntax
| Form | Resolves to |
|---|---|
test-schema |
Named list (exact match) |
21b0 |
UUID list (prefix match) |
test-schema@3 |
Task #3 in named list |
@Scaffold |
First task with "Scaffold" in subject |
| (omit) | Interactive picker |
Task Project Scoping
By default, only task lists belonging to the current project are shown:
- Named lists -- always included
- UUID lists -- included only if the UUID matches a session from
~/.claude/projects/<project-dir>/
Use --all to show all task lists regardless of project.
skills (default: list)
Show the effective outfit across all layers.
shan skills # default: list
shan skills list # explicit
skills on <targets> [--scope user] [--strict] [--fail-on-missing-dependencies]
Turn on one or more skills or groups. Targets are comma-separated colon-syntax names. Missing dependency closures are auto-activated by default.
shan skills on playwright # single skill
shan skills on ts # entire group
shan skills on ts:tooling,playwright,linear # batch
shan skills on --scope user git # user-level
shan skills on dispatch --fail-on-missing-dependencies
skills off [targets] [--scope user] [--strict] [--cascade-dependencies] [--fail-on-dependents]
Turn off skills. No targets = reset all pluggable skills. Dependents cascade by default; dependencies only cascade when requested.
shan skills off playwright # single skill
shan skills off ts # entire group
shan skills off cmux --fail-on-dependents # refuse if active dependents exist
shan skills off dispatch --cascade-dependencies
shan skills off # reset: off ALL pluggable
shan skills off --scope user # reset at user level
skills history [--scope user]
Show the operation log for the current scope.
skills undo [N] [--scope user]
Undo the last N operations (default 1). Restores outfit to pre-operation snapshot.
skills redo [N] [--scope user]
Redo the last N undone operations (default 1).
skills move <axis> <direction> <targets> [--scope user] [--strict] [--cascade-dependencies]
Migrate skills between scopes or commitments. Targets are comma-separated colon-syntax names. Atomic batch: all-or-nothing validation.
Axes:
scope— move between user and project librarycommitment— move between pluggable and core
Directions:
up— scope: project→user, commitment: pluggable→coredown— scope: user→project, commitment: core→pluggable
shan skills move scope down playwright # user library → project library
shan skills move scope up playwright # project library → user library
shan skills move scope up dispatch --cascade-dependencies
shan skills move commitment up playwright # pluggable → core (copy to outfit)
shan skills move commitment down playwright # core → pluggable (auto-install)
shan skills move scope down tips,linear # batch move
Undo/redo fully supported for all move operations via composite history entries.
doctor skills [--no-fix]
Run aspect-based health checks. Default: detect + auto-fix. Use --no-fix for report-only mode.
18 aspects: agent-mirror, broken-symlink (with git rename detection), state-drift, new-leaf, stale-router, orphaned-router, orphaned-scope, stale-gitignore, frontmatter-mismatch, name-conflict, duplicate-name, shadow, stale-shadow, cross-scope-install, dependency-declaration, dependency-cycle, dependency-active-graph.
shan doctor skills # detect + auto-fix
shan doctor skills --no-fix # report only