Skills audited
7
Skills + Policy Audit - IRIS
First-class enterprise capabilities and guardrails for governed agent execution
SkillsPolicy runtimeExecution guardrails
Skills + Policy Framework Audit
This page separates what agents can do from what they are allowed to do. Skills are enterprise capabilities; policies are enterprise guardrails. Every governed decision must prove the skill, authority, policy version, evidence gate, replay path, and human override boundary.
Authorized
1
Review required
0
Not authorized
7
Policy versions
2
Skills Framework
Skills define enterprise capabilities: match invoice, validate journal, explain variance, recommend accrual, assess credit, or optimize payment timing. A skill is not an autonomous right to act; it is a governed capability contract.
| Purpose | Business outcome the skill supports |
|---|---|
| Inputs | Data, documents, events, and context required |
| Outputs | Recommendations or actions produced |
| Authority | Read-only, recommend, approve, or execute |
| Evidence | Proof required before recommendation or execution |
| Explainability | Reasoning and source context available |
| Replay | Execution can be reconstructed later |
| Learning | Outcome feedback can improve future recommendations |
| Human override | Human authority boundary remains explicit |
Policy Framework
Policies define enterprise guardrails: approval limits, segregation of duties, payment thresholds, materiality rules, control requirements, and exception handling.
| Source | Policy Runtime / authority matrix |
|---|---|
| Owner | Finance Governance |
| Scope | Process, domain, case type, and entity |
| Enforcement | Advisory or mandatory |
| Versioning | Historical versions replayable |
| Exceptions | Approved through human governance |
| Evidence requirement | Proof required by policy |
| Monitoring | Compliance measured through ledger and replay |
| Auditability | Every decision traces to policy |
Live Skills + Policy Runtime Proof
Live runtime
Skills audited
7
Authorized
1
Review required
0
Not authorized
7
| Skill | Case | Authority | Policy | Status | Explainable | Replay |
|---|---|---|---|---|---|---|
| Hold Duplicate Payment SKILL-P2P-DUP-HOLD | INV-EKF-L5A-FINAL-20260712 invoice | recommend + controlled execute governed action | invoice_triage_policy v2026.04.15 | authorized | Yes | Yes |
| Assess Counterparty Credit SKILL-CREDIT-EXPOSURE-ASSESS | CREDIT-CP-ATLANTIC-ENERGY counterparty_credit | recommend + escalate recommend + escalate | Policy missing vversion missing | not authorized | Review | Missing |
| Recommend Maintenance Accrual SKILL-CLOSE-MAINT-ACCRUAL | MAINT-ACCRUAL-WO-7781 maintenance_accrual | recommend + controller approval recommend + controller approval | Policy missing vversion missing | not authorized | Review | Missing |
| Assess Trading Exposure SKILL-TRADING-EXPOSURE-ASSESS | TRD-BRENT-FX-EXPOSURE-202606 trade_exposure | recommend + escalate recommend + escalate | Policy missing vversion missing | not authorized | Review | Missing |
| Optimize Payment Timing SKILL-TRSY-PAYMENT-TIMING | TREAS-LIQ-HEADROOM-202606 liquidity_position | recommend + treasury approval recommend + treasury approval | Policy missing vversion missing | not authorized | Review | Missing |
| Explain Forecast Variance SKILL-FPA-VARIANCE-EXPLAIN | FPA-LNG-THROUGHPUT-JUN26 forecast_variance | read-only + recommend read-only + recommend | Policy missing vversion missing | not authorized | Review | Missing |
| Validate Journal SKILL-R2R-JOURNAL-VALIDATE | JE-R2R-ACCRUAL-8421 journal | recommend + approval gated recommend + approval gate | Policy missing vversion missing | not authorized | Review | Missing |
| Hold Duplicate Payment SKILL-P2P-DUP-HOLD | INV-LIVE-DUP-1778594510 invoice | recommend + controlled execute governed action | Policy missing vversion missing | not authorized | Review | Missing |
Runtime relationship
Agents do not execute because they are intelligent. They execute only when an enterprise skill is authorized by policy, supported by evidence, and governed through replay and accountability.
Context
Skill Activated: Hold Duplicate Payment (SKILL-P2P-DUP-HOLD)
Policy Evaluation: invoice_triage_policy
Evidence Collection
Decision Recommendation: REVIEW
Human / Agent Action
Replay & Audit
Learning
Selected skill proof
| Skill owner | P2P Control Owner |
|---|---|
| Risk class | High financial control |
| Evidence gate | Invoice, Purchase order, Receipt, Prior payment |
| Systems impacted | SAP payment block, Supplier validation workflow, Evidence ledger |
| Human override | Available |
Core workshop message
Agents do not execute because they are intelligent. Agents execute because a skill is authorized by policy, supported by evidence, and governed through replay and accountability.