메뉴
Use when a proposal must evidence AI-specific trust and compliance controls inside multi-tenant SaaS, including model providers, evaluation practice, data use, and region routing; use the SaaS trust skill for non-AI credentials.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when reviewing new or changed skills and bundled resources for unsafe installers, credential harvesting, hidden execution, or excessive permissions; use skill-writing for authoring structure.
Use when creating or normalising reusable proposal skills, trigger routes, contracts, references, or authoring automation; use skill-safety-audit instead for a read-only security review.
Use when routing, drafting, reviewing, or assembling a complete consulting proposal, EoI, bid, or tender response; use a numbered pipeline skill when only one known section is required.
Use when choosing whether an AI agent is included, sold as an add-on, or sold standalone; use the pricing skill for rate mechanics and the SLA skill for service commitments.
Use when selecting and adapting agent-specific contract exhibits for a proposal, MSA, or statement of work; use the addendum skill when drafting the full MSA/SLA overlay.
Use when defining intervention credits, abort rights, or refund mechanics for an agent engagement; use the SLA skill for broader credits and the outcome-pricing skill for success fees.
| name | ai-on-saas-compliance-credentials |
| description | Use when a proposal must evidence AI-specific trust and compliance controls inside multi-tenant SaaS, including model providers, evaluation practice, data use, and region routing; use the SaaS trust skill for non-AI credentials. |
| metadata | {"portable":true,"compatible_with":["claude-code","codex"]} |
Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.
saas-trust-and-compliance-credentials-section).| Artefact or fact | Required? | Source/provider | If absent |
|---|---|---|---|
The Trust and Compliance posture from the SaaS skill (saas-trust-and-compliance-credentials-section). | Required | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
| The model-provider list and contractual posture. | Required | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
| The eval discipline (datasets, metrics, refresh cadence, CI integration). | Conditional | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
| The red-team practice (frequency, scope, scorecard). | Conditional | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
| The model-card practice (which features have model cards, what they contain, where they are published). | Conditional | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
| The region-routing capability. | Conditional | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
| The sovereign-AI offer (if any). | Conditional | Buyer, ToR, approved records, or discovery | Mark unavailable, request it, and qualify any affected claim or artefact. |
Stop condition: Stop before asserting scope, compliance, value, acceptance, or readiness when a load-bearing input is missing or contradicted.
Recovery: Record the gap, owner, and next evidence step; then return the narrowest qualified proposal content that remains supportable.
| Artefact | Consumer | Observable acceptance condition |
|---|---|---|
| AI Trust Posture paragraph. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| AI Compliance Map table. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| Model-Card Practice subsection. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| Eval-Harness-in-CI subsection. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| Hallucination SLO subsection. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| Red-Team Practice subsection. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| AI Sub-Processor Disclosure subsection. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
| Region Routing and Sovereign-AI Options subsection. | Proposal evaluator, buyer owner, and delivery team | Content is complete, traceable to named inputs, and usable in its stated proposal section without an unsupported claim. |
For agentic engagements, the Trust and Compliance section gains agent-specific subsections: action audit log practice with 99 % completeness SLA, irreversibility gating with named approvers, intervention SLO, kill-switch architecture and quarterly drill with drill log, agent-specific red-team catalogue (prompt injection via tool output, scope manipulation, memory poisoning, identity confusion, multi-agent collusion), agent identity and impersonation policy, action-scope attestation, multi-agent governance (where applicable), and transparency-to-affected-party UX. Sub-processor disclosure expands to name tool-call APIs as well as model providers. Load ../ai-agent-compliance-credentials/SKILL.md for the agent trust template.
| Evidence | Consumer | Acceptance condition |
|---|---|---|
| Input-and-assumption record | Proposal lead and reviewer | Every load-bearing claim maps to a source, approved assumption, or explicit gap. |
| Decision and review record | Buyer owner and delivery lead | The selected option, rationale, owner, stop condition, and approval status are visible. |
| Section acceptance check | Evaluator-readiness reviewer | Each output meets its stated acceptance condition and unresolved checks are not presented as passed. |
This skill may read supplied tender, discovery, architecture, commercial, security, and operating evidence and draft proposal artefacts within the authorised workspace. It must not publish, send, certify compliance, accept contractual terms, change production systems, spend funds, or disclose confidential evidence without explicit authority. Review and analysis remain read-only by default.
If files, current legal or technical evidence, calculation tools, network access, or reviewers are unavailable, produce the narrowest useful qualified draft. Label assumptions and checks as not assessed, omit unsupported assurances or figures, and state the exact evidence and owner needed to complete the work. An unavailable check never becomes a pass.
| Choice | Action | Failure or risk avoided |
|---|---|---|
| Credential status | Label each control verified, planned, inherited, or not evidenced | Misrepresentation to procurement or regulators |
| Missing load-bearing evidence | Stop the affected claim, record the gap and owner, and continue only with separable supported content. | Fabricated assurance or hidden dependency |
| Conflicting buyer and supplier evidence | Surface the conflict for decision; do not silently choose the favourable version. | Misaligned scope, price, or acceptance |
| Irreversible, contractual, publishing, or production action | Obtain explicit authority and preserve an approval record before acting. | Unauthorised commitment or mutation |
If model provenance or evaluation records cannot be produced, state the gap and current controls, omit the affected trust assurance, and require compliance-owner review before submission.