memory-capture-procedure

name	memory-capture-procedure
description	Capture a numbered implementation algorithm into procedures memory after completing a feature end-to-end. Use when a new feature/system was built — records the exact steps so the next similar build starts from this algorithm.
user-invocable	true

Memory Capture — Procedure

SECURITY — Two-Layer Defence (PCI-DSS 4.0 · ISO/IEC 27001:2022 · GDPR · SOC 2)

Layer 1 (automatic): The hook script pipes all content through pci-scrub.py before sending to the API. Catches: emails, API keys (OpenAI/Anthropic/AWS/GitHub/…), IPv4/IPv6, credit card numbers, national IDs (US SSN, UK NI, EU VAT, Aadhaar, …), connection string credentials, home paths, phone numbers.

Layer 2 (YOU): Regex cannot catch everything. Before calling any memory tool, apply semantic judgment. Do NOT record:

• Person names, usernames, employee IDs → replace with [PERSON]
• Company / customer names → [COMPANY]
• Passwords, MFA codes, session tokens, cookies → [REDACTED]
• Health / medical data (HIPAA PHI) → [REDACTED]
• Financial account numbers, balances → [REDACTED]
• Precise location data (GPS, home address) → [ADDRESS]
• Anything business-confidential (pricing, contracts) → [REDACTED]

When in doubt, leave it out. Over-redaction is always preferable to a compliance violation.

Capture a reusable implementation algorithm into procedures ring memory.

When to Use

• New feature implemented end-to-end (PR merged / push to main)
• New service / integration set up from scratch
• Complex migration or deployment completed
• Multi-step workflow established that will be repeated

What to Capture

Structure as a numbered algorithm with branches:

PROCEDURE: <name> (v1.0)
DOMAIN: <e.g., database-migration, auth-setup, api-integration>

ALGORITHM:
1. <first step>
   - Sub-step a
   - Sub-step b
   ERROR: if X → do Y
2. <second step>
3. <third step>
   BRANCH: if condition A → go to step 5
           if condition B → go to step 4
4. ...
N. VERIFY: <how to confirm the procedure succeeded>

KNOWN PITFALLS:
- <thing that goes wrong and how to avoid it>

CROSS-REFERENCES:
- Related to procedure: <other-procedure-name>
- Files changed: <list key files>

How to Record

1. Write out the full numbered algorithm
2. Scrub PII: replace emails, API keys, paths with [REDACTED] tokens
3. Cross-reference check — before recording, search for similar existing procedures:

   pluggedin_memory_search("<procedure domain> procedure algorithm")
   

   • If similarity > 0.8 with an existing procedure: amend it (add a new version note) rather than creating a duplicate. Use pluggedin_memory_details to fetch the full text of the similar procedure, then record the amended version.
   • If no close match: record as a new procedure (step 4 below).
4. Call pluggedin_memory_observe with:
   • sessionUuid: current session UUID
   • observationType: workflow_step
   • content: the full procedure text
   • outcome: "success"
   • metadata: {"ring": "procedures", "procedure_name": "<name>", "version": "1.0"}

Auto-Trigger Condition

This skill fires automatically when a git push to main / PR merge is detected, or when the user marks a plan step as the final step.

Example

PROCEDURE: central-logging-setup (v1.0)
DOMAIN: observability

ALGORITHM:
1. Install logger package: pnpm add winston
2. Create lib/logger.ts with LoggerService class
   - Configure transports: Console (dev), File (prod)
   - Export singleton instance
3. Update all existing service files:
   - Import logger from lib/logger
   - Replace console.log → logger.info
   - Replace console.error → logger.error
   ERROR: if file uses 'use client' → skip (client components can't use Node logger)
4. Add request logging middleware to app/api/ routes
5. VERIFY: pnpm build passes, logs appear in /var/log/app/

KNOWN PITFALLS:
- winston is not Edge-runtime compatible — exclude from middleware bundle
- File transport needs write permission on /var/log/app/

CROSS-REFERENCES:
- Any new service function added to lib/ must call logger per this procedure

---

Skip Guidance

Skip this skill if:

• The information is already obvious from the current conversation context
• The content contains personal names, customer data, or business-confidential information that cannot be fully redacted
• The observation is trivial (e.g., ran ls, checked a file, viewed a README)
• A <private> block wraps the relevant content — respect user opt-out