pi-share_skill

This skill loads and decodes pi-share session transcripts to extract full conversations and summarize human actions for analysis.
  • TypeScript

705

GitHub Stars

2

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill mitsuhiko/agent-stuff --skill pi-share

  • fetch-session.mjs9.8 KB
  • SKILL.md3.7 KB

Overview

This skill loads and parses session transcripts exported from pi-share URLs (shittycodingagent.ai / buildwithpi.ai / buildwithpi.com). It fetches the underlying GitHub Gist, decodes the embedded session HTML, and extracts the conversation history and metadata. Use it to inspect full sessions, get compact JSON entries, or produce human-focused summaries of the user’s behavior.

How this skill works

The script locates a gist ID in the provided URL or accepts a raw gist ID. It fetches the gist, finds the session.html file, and base64-decodes the embedded session payload. The decoded data is parsed into a SessionData structure with header, entries, system prompt, tools, and leafId. Flags let you output the header, entries (JSON lines), system prompt, tool list, or a human-centric summary.

When to use it

  • You have a pi-share/session URL or gist ID and need the full conversation transcript.
  • You want structured JSON entries for programmatic analysis or jq filtering.
  • You need the system prompt or tool definitions used in a session.
  • You want a human-centric summary focusing on what the user did and how they steered the agent.
  • You want to count tool calls, find user messages, or inspect model/thinking changes.

Best practices

  • Provide the full session URL or just the gist ID to avoid parsing errors.
  • Use --entries for machine processing and piping into jq or other tools.
  • Use --header to quickly see session metadata (timestamp, session id, cwd).
  • Run --human-summary when you need a concise, human-focused narrative of the user’s behavior.
  • Validate network access to GitHub API if gists are private or rate-limited.

Example use cases

  • Load a session and filter to user messages: fetch-session.mjs <id> --entries | jq 'select(.type=="message" and .message.role=="user")'
  • Generate a human-centric summary to understand how a user iterated on prompts: fetch-session.mjs <id> --human-summary
  • Extract system prompt and tools to reproduce the agent environment: fetch-session.mjs <id> --system --tools

FAQ

The script accepts full pi-share session URLs containing a gist ID or a raw gist ID string.

What does the human summary emphasize?

The human summary focuses on the user’s initial goal, reprompts and steering, corrections or interventions, and overall prompting style rather than the agent’s outputs.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational