/approve-lab
Approve lab submission to Adaptyv Bio (triple-gated)
$ npx -y skills add 001TMF/blatant-why --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/approve-lab
Context preview
What this command does when you run it.
Approve lab submission to Adaptyv Bio (triple-gated)
Command definition
approve-lab.mdname: by:approve-lab
description: Approve lab submission to Adaptyv Bio (triple-gated)
/approve-lab — Approve Lab Submission (Triple-Gated)
Initiate the lab submission approval process for sending designs to Adaptyv Bio for experimental validation. This is a GATED operation requiring explicit confirmation.
Instructions
Step 0: Read model profile
MODEL_PROFILE=$(cat .by/config.json 2>/dev/null | grep -o '"model_profile"[[:space:]]*:[[:space:]]*"[^"]*"' | grep -o '"[^"]*"$' | tr -d '"' || echo "balanced")
Model lookup for this command: | Agent | quality | balanced | budget | |-------|---------|----------|--------| | by-lab | opus | opus | sonnet |
Step 1: Pre-flight checks
Verify before proceeding:
- An active campaign exists with completed screening
- At least 1 design has PASS status in screening results
- Environment has Adaptyv API key configured
- Campaign state is at RANKING phase or later
If any check fails, report the issue and stop. Do NOT proceed.
Step 2: Show submission summary
Display what will be submitted:
- Campaign ID and target name
- Number of designs to submit (PASS status only)
- Top designs with scores (ipSAE, ipTM, p_bind)
- Estimated cost and turnaround time
- Adaptyv Bio endpoint being used
Step 3: Require explicit confirmation
Ask the user to type **CONFIRM** to proceed.
If the user does not type CONFIRM, abort with: "Lab submission cancelled. No data was sent."
Step 4: Write approval file (Layer 3)
cat > .by/campaigns/$CAMPAIGN_ID/lab/approval.json << EOF
{
"approved": true,
"approved_by": "user",
"approved_at": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
"ttl_minutes": 5,
"campaign_id": "$CAMPAIGN_ID",
"design_count": $DESIGN_COUNT
}
EOFStep 5: Update campaign state (Layer 2)
Set `labApproved: true` in `state.json`.
Step 6: Spawn by-lab agent
Delegate to a **by-lab** agent (model per profile table above):
> Submit approved designs from `{campaign_dir}` to Adaptyv Bio. > Use the MCP confirmation code from the adaptyv server (Layer 1). > Verify approval.json exists and TTL has not expired (5 minutes). > Verify campaignState.labApproved is true. > Log all submission details to `{campaign_dir}/logs/lab_submission.log`.
Step 7: Report outcome
Show submission confirmation or failure details. Include order ID and expected turnaround if successful.
**CRITICAL: NEVER bypass any of the three approval layers.**
Read more
name: by:approve-lab description: Approve lab submission to Adaptyv Bio (triple-gated)
/approve-lab — Approve Lab Submission (Triple-Gated)
Initiate the lab submission approval process for sending designs to Adaptyv Bio for experimental validation. This is a GATED operation requiring explicit confirmation.
Instructions
Step 0: Read model profile
MODEL_PROFILE=$(cat .by/config.json 2>/dev/null | grep -o '"model_profile"[[:space:]]*:[[:space:]]*"[^"]*"' | grep -o '"[^"]*"$' | tr -d '"' || echo "balanced")
Model lookup for this command: | Agent | quality | balanced | budget | |-------|---------|----------|--------| | by-lab | opus | opus | sonnet |
Step 1: Pre-flight checks
Verify before proceeding:
- An active campaign exists with completed screening
- At least 1 design has PASS status in screening results
- Environment has Adaptyv API key configured
- Campaign state is at RANKING phase or later
If any check fails, report the issue and stop. Do NOT proceed.
Step 2: Show submission summary
Display what will be submitted:
- Campaign ID and target name
- Number of designs to submit (PASS status only)
- Top designs with scores (ipSAE, ipTM, p_bind)
- Estimated cost and turnaround time
- Adaptyv Bio endpoint being used
Step 3: Require explicit confirmation
Ask the user to type **CONFIRM** to proceed.
If the user does not type CONFIRM, abort with: "Lab submission cancelled. No data was sent."
Step 4: Write approval file (Layer 3)
cat > .by/campaigns/$CAMPAIGN_ID/lab/approval.json << EOF
{
"approved": true,
"approved_by": "user",
"approved_at": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
"ttl_minutes": 5,
"campaign_id": "$CAMPAIGN_ID",
"design_count": $DESIGN_COUNT
}
EOFStep 5: Update campaign state (Layer 2)
Set `labApproved: true` in `state.json`.
Step 6: Spawn by-lab agent
Delegate to a **by-lab** agent (model per profile table above):
> Submit approved designs from `{campaign_dir}` to Adaptyv Bio. > Use the MCP confirmation code from the adaptyv server (Layer 1). > Verify approval.json exists and TTL has not expired (5 minutes). > Verify campaignState.labApproved is true. > Log all submission details to `{campaign_dir}/logs/lab_submission.log`.
Step 7: Report outcome
Show submission confirmation or failure details. Include order ID and expected turnaround if successful.
**CRITICAL: NEVER bypass any of the three approval layers.**
AI-powered biologics design campaign agent — multi-agent orchestration with BoltzGen, PXDesign, Protenix, and 200+ cloud tools. Antibodies, nanobodies, de novo binders, and beyond.

