pi-share — Claude Code skill: what it does, how to install
pi-share loads shared pi sessions and parses their conversation history. Use it to inspect messages, tool calls, or the human’s actions.
Apache-2.0 125 installs ★ 3,224 updated 2026-09-27
npx skills add mitsuhiko/agent-stuff --skill pi-share -a claude-codeWhat pi-share does
pi-share accepts a session URL or gist ID, fetches the exported data from a GitHub Gist, and extracts the conversation. The parsed session can include its header, messages, model changes, thinking-level changes, and context compaction events.
The skill includes fetch-session.mjs. It can print the complete session, the header alone, JSON Lines entries, the system prompt, or tool definitions. With --human-summary, it produces a human-focused account of goals, follow-up prompts, corrections, interventions, and prompting style.
The --human-summary mode calls claude-haiku-4-5 through pi -p, so pi must be available for that report. Other output modes use the helper script directly.
What it helps with
- load a session from a gist ID
- extract messages and tool calls
- inspect the exported system prompt
- print the available tool definitions
- summarize how a person steered a session
Example prompts for the agent
- Load the complete session history for the gist ID I provided.
- Print the session entries as JSON Lines and filter them to user messages.
- Create a human-focused session summary with
--human-summary.
How to install pi-share
1Skills CLI (Vercel)
npx skills add mitsuhiko/agent-stuff --skill pi-share -a claude-code2Manually: put the files in place
~/.claude/skills/pi-share/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/pi-share/
1Skills CLI (Vercel)
npx skills add mitsuhiko/agent-stuff --skill pi-share -a codex2Manually: put the files in place
.agents/skills/pi-share/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add mitsuhiko/agent-stuff --skill pi-share -a cursor2Manually: put the files in place
.cursor/skills/pi-share/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install mitsuhiko/agent-stuff skills/pi-share2Manually: put the files in place
.github/skills/pi-share/SKILL.md1Skills CLI (Vercel)
npx skills add mitsuhiko/agent-stuff --skill pi-share -a gemini-cli2Platform CLI
gemini skills install https://github.com/mitsuhiko/agent-stuff --path skills/pi-shareManually: put the files in place
.devin/skills/pi-share/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add mitsuhiko/agent-stuff --skill pi-share -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about pi-share
How do I install pi-share in Claude Code?
Run npx skills add mitsuhiko/agent-stuff --skill pi-share -a claude-code, then check ~/.claude/skills/pi-share/SKILL.md.
What does pi-share do?
It fetches a session export from a GitHub Gist, decodes it, and exposes the conversation history. Separate modes can print the header, entries, system prompt, or tool definitions.
What does --human-summary require?
That mode analyzes a condensed transcript through pi -p with claude-haiku-4-5. The pi command must be available.
Source text
SKILL.md · Apache-2.0
---
name: pi-share
description: "Load and parse session transcripts from shittycodingagent.ai/buildwithpi.ai/buildwithpi.com/pi.dev (pi-share) URLs. Fetches gists, decodes embedded session data, and extracts conversation history."
---
# pi-share / buildwithpi Session Loader
Load and parse session transcripts from pi-share URLs (shittycodingagent.ai, buildwithpi.ai, buildwithpi.com, pi.dev).
## When to Use
**Loading sessions:** Use this skill when the user provides a URL like:
- `https://shittycodingagent.ai/session/?<gist_id>`
- `https://buildwithpi.ai/session/?<gist_id>`
- `https://buildwithpi.com/session/?<gist_id>`
- `https://pi.dev/session/?<gist_id>`
- `https://pi.dev/session/#<gist_id>`
- Or just a gist ID like `46aee35206aefe99257bc5d5e60c6121`
- Or hash-prefixed shorthand like `#46aee35206aefe99257bc5d5e60c6121`
**Human summaries:** Use `--human-summary` when the user asks you to:
- Summarize what a human did in a pi/coding agent session
- Understand how a user interacted with an agent
- Analyze user behavior, steering patterns, or prompting style
- Get a human-centric view of a session (not what the agent did, but what the human did)
The human summary focuses on: initial goals, re-prompts, steering/corrections, interventions, and overall prompting style.
## How It Works
1. Session exports are stored as GitHub Gists
2. The URL contains a gist ID after the `?`
3. The gist contains a `session.html` file with base64-encoded session data
4. The helper script fetches and decodes this to extract the full conversation
## Usage
```bash
# Get full session data (default)
node ~/.pi/agent/skills/pi-share/fetch-session.mjs "<url-or-gist-id>"
# Get just the header
node ~/.pi/agent/skills/pi-share/fetch-session.mjs <gist-id> --header
# Get entries as JSON lines (one entry per line)
node ~/.pi/agent/skills/pi-share/fetch-session.mjs <gist-id> --entries
# Get the system prompt
node ~/.pi/agent/skills/pi-share/fetch-session.mjs <gist-id> --system
# Get tool definitions
node ~/.pi/agent/skills/pi-share/fetch-session.mjs <gist-id> --tools
# Get human-centric summary (what did the human do in this session?)
node ~/.pi/agent/skills/pi-share/fetch-session.mjs <gist-id> --human-summary
```
## Human Summary
The `--human-summary` flag generates a ~300 word summary focused on the human's experience:
- What was their initial goal?
- How often did they re-prompt or steer the agent?
- What kind of interventions did they make? (corrections, clarifications, frustration)
- How specific or vague were their instructions?
This uses claude-haiku-4-5 via `pi -p` to analyze the condensed session transcript.
## Session Data Structure
The decoded session contains:
```typescript
interface SessionData {
header: {
type: "session";
version: number;
id: string; // Session UUID
timestamp: string; // ISO timestamp
cwd: string; // Working directory
};
entries: SessionEntry[]; // Conversation entries (JSON lines format)
leafId: string | null; // Current branch leaf
systemPrompt?: string; // System prompt text
tools?: { name: string; description: string }[];
}
```
Entry types include:
- `message` - User/assistant/toolResult messages with content blocks
- `model_change` - Model switches
- `thinking_level_change` - Thinking mode changes
- `compaction` - Context compaction events
Message content block types:
- `text` - Text content
- `toolCall` - Tool invocation with `toolName` and `args`
- `thinking` - Model thinking content
- `image` - Embedded images
## Example: Analyze a Session
```bash
# Pipe entries through jq to filter
node ~/.pi/agent/skills/pi-share/fetch-session.mjs "<url>" --entries | jq 'select(.type == "message" and .message.role == "user")'
# Count tool calls
node ~/.pi/agent/skills/pi-share/fetch-session.mjs "<url>" --entries | jq -s '[.[] | select(.type == "message") | .message.content[]? | select(.type == "toolCall")] | length'
```
© mitsuhiko · Apache-2.0 · mitsuhiko/agent-stuff