⚡ EXECUTION INSTRUCTIONS FOR CLAUDE

When this command is invoked, YOU (Claude) must execute these steps immediately. This is NOT documentation – run every step below.

🚨 EXECUTION WORKFLOW

1. Gather the testing methodology, evidence summary, and notes for the current work item.
2. Identify any artifacts (logs, screenshots, DOM dumps, JSON, etc.) that should be preserved.
3. Execute the /savetmp command exactly as documented to store everything under /tmp/<repo>/<branch>/<work>/.
4. Confirm the command output and record the generated path in the conversation.

📋 REFERENCE DOCUMENTATION

/savetmp – Structured Evidence Archiver

Purpose: Capture testing methodology, evidence summaries, and supporting artifacts in a predictable /tmp evidence package scoped by repo → branch → work item.

Usage: /savetmp <work_name> [options]

Primary Script: .claude/commands/savetmp.py

---

🗂️ Directory Layout

Running /savetmp creates a directory using this structure:

/tmp/<repo_name>/<branch>/<work_name>/<timestamp>/
  ├── methodology.md   # Testing methodology details (if provided)
  ├── evidence.md      # Evidence summary
  ├── notes.md         # Extra context / TODOs
  ├── artifacts/       # Copied files/directories passed via --artifact
  ├── metadata.json    # Machine-readable summary of the run
  └── README.md        # Quick overview + artifact list

• <repo_name> is derived from the git root (fallback: current directory name).
• <branch> comes from git rev-parse --abbrev-ref HEAD (fallback: detached).
• <work_name> is provided when invoking the command (e.g., sidebar-regression-2025-01-15).
• <timestamp> is in UTC (YYYYMMDDTHHMMSSZ) to keep multiple runs without collision.

---

🛠️ Command Syntax

python .claude/commands/savetmp.py "<work_name>" \
  --methodology "<short description>" \
  --methodology-file <path/to/detailed-methodology.md> \
  --evidence "<evidence summary>" \
  --evidence-file <path/to/evidence-details.md> \
  --notes "<follow-up notes>" \
  --notes-file <path/to/extra-notes.txt> \
  --artifact <path/to/file-or-directory> \
  --artifact <path/to/another-artifact>

Flags

Flag	Description
--methodology	Inline description of the testing methodology used.
--methodology-file	Append contents of a file (e.g., generated test plan).
--evidence	Summary of results or notable findings.
--evidence-file	Append detailed evidence from a file.
--notes	Additional notes, TODOs, or observations.
--notes-file	Append extra context from a file.
--artifact	Copy a file or directory into artifacts/. Use multiple times for multiple artifacts.

• All --*-file flags are optional; missing files are skipped with a warning.
• Artifact paths can be absolute or relative; directories are copied recursively.

---

✅ Step-by-Step Execution Template

1. Compose methodology text (optional but recommended):

   cat <<'EOF' > /tmp/methodology.txt
   - Ran npm test -- --run
   - Executed browser MCP scenario: conversation-management-flow-test.md
   - Verified sidebar toggle in both modes manually
   EOF
   

2. Prepare evidence summary:

   cat <<'EOF' > /tmp/evidence.txt
   - Vitest suite: 101/101 passing
   - Browser MCP artifacts saved under ./testing_llm/evidence/convo/
   - No console errors observed
   EOF
   

3. Collect artifacts (copy or point to existing directories):
   • Example: /tmp/run_logs/convo/, testing_llm/evidence/convo/sidebar/
4. Invoke the command:

   python .claude/commands/savetmp.py "conversation-regression-jan15" \
     --methodology-file /tmp/methodology.txt \
     --evidence-file /tmp/evidence.txt \
     --notes "Ready for PR evidence upload" \
     --artifact /tmp/run_logs/convo \
     --artifact testing_llm/evidence/convo/sidebar
   

5. Record the output: The script prints both the base directory and the precise run folder. Paste this path back into the conversation or task log.

---

🧪 Verification Checklist

• Command exits with status 0 and prints ✅ Evidence archived.
• metadata.json contains the repo, branch, work name, and artifact list.
• README.md enumerates any saved sections (methodology/evidence/notes).
• Artifacts appear under artifacts/ with expected filenames.
• Timestamped subdirectory created under /tmp/<repo>/<branch>/<work_name>/.

If any step fails, capture stderr/stdout and re-run after fixing inputs.

---

🔁 Re-running / Updating Evidence

• Re-invoke /savetmp with the same work_name; a new timestamped subdirectory is created.
• To update only notes/evidence, you can omit --artifact (artifacts are only copied when provided).
• The most recent run path is printed at the end—share it when handing off the work item or writing the PR summary.

---

🧹 Cleanup (Optional)

Evidence stays under /tmp; remove stale runs if needed:

rm -rf /tmp/<repo>/<branch>/<work_name>/<timestamp>

(Only delete directories you created; never touch shared evidence owned by other engineers without confirmation.)