session-dump

name

session-dump

description

Generate a continuation-ready markdown dump for an external coding-agent session, starting with Claude Code session IDs. Use when the user needs to continue prior work from a Claude session, inspect where session history is stored, dump tool calls/subagents/tasks/plans/teams metadata, or build a citation-heavy session handoff. The bundled Bun TypeScript CLI runs standalone and supports reference-first output with opt-in raw inclusion.

Session Dump

Use this skill when a user gives you a session id from another agent and wants a full, citation-heavy dump that another agent can use to continue work.

The bundled CLI is Bun-first TypeScript and currently supports:

• claude provider
• reference-first markdown dumps by default
• opt-in raw inclusion for referenced artifacts
• built-in heuristic redaction with no third-party runtime dependencies

Entry point

From the skill directory, run:

bun run session-dump --provider claude --session <session-id>

Or directly:

bun run ./src/index.ts --provider claude --session <session-id>

Defaults

• Main transcript is inlined
• Related artifacts are cited by absolute path
• Raw referenced file content is not embedded unless requested
• Redaction is enabled by default

Useful flags

• --output <file>: write the markdown dump to a file
• --include-raw: inline all discovered related artifacts
• --include-subagents: inline subagent transcripts
• --include-tool-results: inline spilled tool-result files
• --include-debug: inline debug log content
• --include-history: inline matching history.jsonl entries
• --include-tasks: inline task JSON files
• --include-teams: inline team config and inbox files
• --include-plans: inline plan files
• --no-redact: disable heuristic redaction
• --json-index <file>: write a machine-readable artifact manifest
• --strict: fail when a referenced artifact is missing

Claude artifact model

The Claude adapter checks these locations:

• ~/.claude/projects/*/<session-id>.jsonl for the canonical transcript
• sibling subagents/ and tool-results/
• ~/.claude/debug/<session-id>.txt
• ~/.claude/history.jsonl
• ~/.claude/tasks/<session-id>/
• ~/.claude/teams/*/config.json and inboxes when linked by leadSessionId
• ~/.claude/plans/*.md from explicit plan references and slug matches
• ~/.claude/session-env/ opportunistically

Output contract

The dump is deterministic and markdown-only in v1. It includes:

• session identity and metadata
• artifact inventory with absolute paths
• main transcript rendering
• related artifact sections
• explicit missing-data notes
• non-LLM continuation notes derived from transcript structure

Testing

Use Bun’s built-in test runner:

bun test
bun test --coverage

If you update parsing or redaction behavior, run coverage and keep tests focused on:

• transcript discovery
• persisted-output path extraction
• redaction heuristics
• artifact rendering behavior