hashgraph-online/awesome-codex-plugins / plugins/AltByteSG/personal-data-protection-skill/skills/personal-data-protection
personal-data-protection
VerifiedPersonal-data-protection compliance reference for engineers building applications subject to Singapore PDPA, Indonesia UU PDP, Thailand PDPA, Malaysia PDPA, Philippines DPA (RA 10173), or Vietnam PDPL (91/2025/QH15). Use when reviewing or modifying code that touches personal data — signup/auth/consent, data export, account deletion, retention/purging, admin access to personal-data stores, third-party processors, breach response, or privacy/T&C documents.
layers/04-controls-and-processes.md
View on GitHub# Layer 04 — Controls and Processes
> ⚠ **Reference material only — not legal advice.** See [DISCLAIMER.md](https://github.com/AltByteSG/personal-data-protection-skill/blob/main/DISCLAIMER.md). Verify against the official statute and consult a qualified DPO / lawyer.
Server-side and pipeline code that enforces the obligations. Universal patterns across stacks.
## Contents
- User-rights endpoints
- Admin endpoint pattern
- Retention sweeps
- Storage cleanup orchestration
- Sub-processor dispatch (push / email / webhooks)
- Logging hygiene patterns
- Breach detection signals
- Session and credential rotation
- Backwards compatibility for personal-data changes
## User-rights endpoints
Every PDP-family statute requires support for at least:
- **Access** — the user can request a copy of all personal data the organisation holds about them.
- **Correction** — the user can request rectification of inaccurate or incomplete data.
- **Deletion** — the user can request deletion of their personal data.
- **Withdrawal of consent** — the user can revoke any consent previously given.
Implementation suggestions:
| Right | Common implementation |
|---|---|
| Access | A function or endpoint that returns the user's complete personal-data set as a structured payload (typically JSON), delivered to the user via in-app share-sheet or download link. Rate-limit to prevent abuse (e.g. once per 7 days). |
| Correction | Self-service edit screens for fields the user can correct themselves, plus a contact channel for fields under admin control. |
| Deletion | Hard-delete cascade for owned data; SET NULL or "Deleted User" attribution for shared content. Triple-confirm in the UI to prevent accidents. Re-authenticate before delete. |
| Withdrawal | The same toggles used to grant consent must allow revocation, with equal effort. |
**Access endpoint completeness** is critical and frequently fails audits. The implementation must return *all* personal data the org holds — not just the most-used fields. See [`checklists/new-data-field.md`](../checklists/new-data-field.md) for the discipline of keeping the export in sync as new fields are added.
## Admin endpoint pattern
Every admin endpoint (read or write) should follow the same shape:
1. **Verify the caller has the admin role** at the start of the function. Don't trust middleware alone — check inside the function too. Defence in depth.
2. **Mutations** call the audit-log helper at the end, recording action, resource type, resource ID, and a JSON details blob.
3. **Reads of single-user PII or PII lists** call a throttled audit-log helper at the end (5-minute deduplication per actor + resource).
4. **Reads of pure aggregates** (dashboard, KPIs, charts) skip the audit log — document the exclusion in a code comment.
Why throttle reads but not writes: every mutation is unique state change. Reads can repeat (page refreshes, polling). Without throttling, a user-detail page refresh produces dozens of identical audit rows that drown the signal.
## Retention sweeps
Statutes require ceasing to retain personal data once the purpose is no longer served. For most categories this means scheduled background cleanup.
**Pattern:**
- One canonical "PDP cleanup" function that runs daily at off-peak hours.
- Each retention rule lives as a separate sub-function called from the canonical one (`cleanup_expired_messages()`, `cleanup_expired_events()`, `cleanup_orphan_storage()`, etc.).
- The schedule is registered with whatever cron / scheduler your platform offers.
- Each sub-function is idempotent — safe to run again if a previous run failed mid-way.
**When adding a new retention rule:**
1. Write the cleanup sub-function (parameterised by the retention period).
2. Add a call to it from the canonical sweep.
3. Document the rule in the privacy policy retention table.
4. Don't create a new schedule — reuse the existing one.
**File / object storage cleanup** has a special caveat on most platforms: deleting database rows does NOT automatically delete object-storage files. Cleanup must explicitly call the storage API for each file, in addition to the database delete.
## Storage cleanup orchestration
When deleting account-owned files from object storage:
1. Use the storage platform's REST API or SDK with service credentials.
2. Batch deletes (most APIs accept a list per call).
3. Tolerate partial failures — the next sweep run should re-attempt.
4. **Do not** rely on client-side cleanup as the primary path. The server is authoritative; client cleanup is at most a UX optimisation.
For orphan reaping (files whose owner no longer exists in the database):
- Periodic full scan of bucket prefixes.
- For each file, extract the owner identifier from the path.
- LEFT JOIN against the users table; delete files whose owner is NULL.
- Skip prefixes that are conversation-keyed or otherwise not owned by an individual user (group chat media, shared content).
**Shared containers need different cleanup logic than user-owned files.** If a file lives in a shared path (group chat media, collaborative documents, event galleries), deleting it on individual account deletion would erase data the *other* members are still entitled to view. Pattern: re-attribute ownership to a system "Deleted User" pseudo-account on account deletion; reap the file itself only when *all* referencing members are gone or when the parent container (the conversation, the event, the document) is itself deleted. The trade-off: a deleted user's content lingers in shared spaces — name them in the privacy policy if you want this to be defensible.
## Sub-processor dispatch (push / email / webhooks)
When your system dispatches data to a sub-processor (push notification provider, email service, analytics, webhooks):
- **Authenticate** with credentials from your secret store, never hard-coded.
- **Sanitise the payload** so it contains the minimum data needed — the destination provider sees what you send.
- **Cache short-lived auth tokens** with sensible expiry; don't re-acquire on every call.
- **Handle stale credentials** (expired tokens, revoked keys) gracefully — clear them and log without leaking the credential value.
- **Sanitise the response** before logging — sub-processors notoriously echo the value you submitted (email address, push token, OAuth token) back in error responses.
## Logging hygiene patterns
For every `console.error` / `console.log` / `print` / equivalent in server-side code that includes a variable, the variable must be sanitised first. Suggested helpers (translate to your language):
```
// Pseudocode — adapt to your language
function safeError(e) {
// Extract code + message; never dump full object
return e.code ? `${e.code}: ${e.message}` : e.message || String(e);
}
function redactId(uuid) {
// First 8 chars + ellipsis. Enough to correlate, not enough to identify
return uuid ? `${uuid.slice(0, 8)}…` : '<none>';
}
function redactPII(s) {
// Multi-pass: named token fields, emails, then long token-shaped substrings
let out = s.slice(0, 300);
out = out.replace(/(token|access_token|refresh_token|authorization|api_key)\s*[:=]\s*['"]?([^'",\s}]{8,})/gi, '$1: <redacted>');
out = out.replace(/[\w.+-]+@[\w.-]+\.\w+/g, '<email>');
out = out.replace(/\b[A-Za-z0-9][A-Za-z0-9._:/=+-]{39,}\b/g, '<redacted>');
return out;
}
```
Apply at every call site that logs a variable. Plain string literals are fine.
## Breach detection signals
Without some kind of detection, you only learn about breaches when something visible breaks. For the regulatory clock to start (PDP-family typically 3–72 hours after assessment), someone has to notice. Minimum signals to consider:
- **Failed-auth spike alarm** (sudden surge in 4xx auth responses)
- **Mass-export detector** (rate-limit alert on data-export endpoint per hour)
- **Storage egress anomaly** (unusual download volume)
- **Repository secret-scanning** (your VCS provider almost certainly offers it; turn it on)
- **Log review schedule** (quarterly skim of admin audit log for unfamiliar actions)
These are operational signals, not user-facing features — owned by [layer 07](07-operational.md).
## Session and credential rotation
When a user changes their password, deletes their account, or revokes a session: existing sessions on other devices should be invalidated. Most auth platforms expose a "sign out everywhere" or "revoke refresh tokens" function — wire it into your account-management flows.
For admin sessions: shorter session lifetimes than user sessions, plus periodic forced re-authentication. The blast radius of an admin-session leak is much higher.
## Backwards compatibility for personal-data changes
The rules are in [layer 02](02-architecture.md#backwards-compatibility-when-changing-personal-data-handling) — they apply unchanged to any function user clients call. Two additions specific to this layer:
- Adding new **optional parameters with defaults** is safe, same as adding return-shape fields.
- For internal admin functions, client version compatibility doesn't apply, but the admin console must be redeployed in the same release.
See also CLAUDE.md golden rule #12 in your project (or whatever its equivalent is) for project-specific backwards-compat policy.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |