PII redaction gate

Scans outbound content — messages, exports, attachments, drafts leaving trusted boundaries — for PII/PHI and blocks flagged items pending human approval, with inline redaction suggestions. Use when the user mentions PII protection, data loss prevention, redaction, outbound compliance, or safe data sharing with external parties.

Every company's worst data leak is one helpful employee attaching one convenient export. Policy documents don't stop it; a gate does. This skill sits on the configured outbound paths, catches personal data before it crosses the boundary, and makes the safe path the fast path: here's the flagged content, here's the redacted version, one click to release.

When to run

Continuous on configured outbound channels: external email, export downloads, external shares, and agent-generated external content (recommended)
On demand: "scan this doc before I send it to the vendor"

Computer scopes

Outbound interception points as configured by the admin (the gate perimeter)
Docs: the org's data-classification policy (what counts, at what sensitivity, for which destinations — the rulebook)
Audit log: every scan, block, release, and override (write)

Workflow

Scan outbound items against the classification rulebook: direct identifiers (names tied to identifying context, contact details, government IDs, financial and health data) and quasi-identifier combinations that identify in aggregate.
Judge by destination: the same record may be fine to a processor under DPA and blocked to an unknown domain — destination tier is part of the rule, not an afterthought.
On flag: block at L2 (automatic, audited, reversible by release), notify the sender with the specific findings highlighted in place — never a vague "policy violation" bounce.
Offer the redacted alternative: same content, findings masked or generalized, ready to send. Release of the original requires L1 human approval by the sender's designated approver, with reason captured.
Log everything; weekly, report gate metrics and top friction points to the compliance owner — a gate that's constantly overridden is a rulebook bug, and the report says so.

Guardrails

Autonomy: L2 to block, L1 to release. The gate can stop content on its own; only a human can wave flagged content through, and every override is logged with reason.
False-positive humility: senders can contest; contested patterns route to rulebook review.
The gate reads outbound content solely for scanning; scan excerpts in notifications go only to the sender and approver.

Output

Per interception: sender notification with highlighted findings + redacted alternative + release path. Weekly: gate metrics (catches, releases, contest rate, friction points).

Eval fixtures

Set: 250 outbound items across channels — 60 with seeded PII/PHI of varying subtlety (including 15 quasi-identifier combinations), 20 destination-dependent cases, 170 clean.
Rubric: catch recall on seeded items; false-positive rate on clean set; destination logic correctness; redacted alternatives preserve utility (human-graded).
Ship gate: recall ≥ 0.95 (1.0 on direct identifiers); false-positive rate ≤ 0.05.

Run pii redaction gate on your own data.

MORE SKILLS

Governance, Security & Compliance · L0

Agent audit reporter

Produces the periodic governance digest of all agent and skill activity — actions taken by autonomy level, approvals granted and by whom, guardrail triggers, denials, and anomalies — from GStack's audit logs. Use when the user mentions AI governance reporting, agent audit, "what did the agents do this week," compliance review of automation, or board/CISO reporting on AI usage.

Governance, Security & Compliance · L0

Agent audit reporter

Produces the periodic governance digest of all agent and skill activity — actions taken by autonomy level, approvals granted and by whom, guardrail triggers, denials, and anomalies — from GStack's audit logs. Use when the user mentions AI governance reporting, agent audit, "what did the agents do this week," compliance review of automation, or board/CISO reporting on AI usage.

Governance, Security & Compliance · L1

SOC 2 evidence collector

Gathers audit evidence for SOC 2 and similar control frameworks from connected systems — access reviews, change logs, backup confirmations, training records — mapped to the control checklist with timestamps and provenance. Use when the user mentions SOC 2, audit prep, evidence collection, control testing, compliance audit season, or an auditor request list.

Governance, Security & Compliance · L1

SOC 2 evidence collector

Gathers audit evidence for SOC 2 and similar control frameworks from connected systems — access reviews, change logs, backup confirmations, training records — mapped to the control checklist with timestamps and provenance. Use when the user mentions SOC 2, audit prep, evidence collection, control testing, compliance audit season, or an auditor request list.