| name | session-history-metadata |
| description | Extract Cursor CLI coding agent harness version and exact token usage for session_history entries. Use when preparing session_history files, filling Agent harness or Token usage sections, or when the user asks how to record version and tokens accurately. |
Session History Metadata
Use This Skill
Use this skill when you need accurate session_history/ metadata for a Cursor CLI coding agent run.
Prefer exact token counts from a saved agent --print --output-format stream-json log or a terminal capture containing that JSON output. The repo utility script also calls agent about --format json to get the CLI version.
Exact Metadata Workflow
-
Capture or locate a saved Cursor CLI stream log.
Example capture:
agent --print --output-format stream-json --trust "Your prompt here" > tmp-agent-stream.jsonl
You can also point the utility at a terminal capture file that contains the same JSON lines.
-
Run the repo utility:
uv run python scripts/session_history_metadata.py --stream-json tmp-agent-stream.jsonl --format markdown --require-usage
-
Paste the emitted markdown block into the ## Agent harness, ## Model, and ## Token usage sections of your session_history/ file.
Notes
-
Default harness label: Cursor CLI coding agent
-
Exact token usage requires a result.usage block from Cursor CLI stream-json output.
-
If you do not have a saved stream log, you can still get the harness version with:
agent about --format json
-
Raw Cursor transcript exports under agent-transcripts/ do not currently provide exact token usage, so do not treat them as an exact source for those numbers.
-
Do not invent token numbers. If the current harness/workflow does not expose them yet, add or update a repo-local utility/skill rather than writing placeholders as if they were exact.
