How do I install the Mandate 2 2 2 Runtime Evidence Intake Guide skill?

Run npx skillmds@latest add jefflyt/mandate-2-2-2-runtime-evidence-intake-guide in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Mandate 2 2 2 Runtime Evidence Intake Guide safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Mandate 2 2 2 Runtime Evidence Intake Guide?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Mandate 2 2 2 Runtime Evidence Intake Guide free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Mandate 2 2 2 Runtime Evidence Intake Guide?

jefflyt (@jefflyt) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Mandate 2 2 2 Runtime Evidence Intake Guide by jefflyt

Name: Mandate 2 2 2 Runtime Evidence Intake Guide
Rating: 4.5 (4 reviews)
Author: jefflyt

Mandate 2 2 2 Runtime Evidence Intake Guide

Gather natural-language context and generate runtime/process evidence collection guidance for Mandate 2.2.2 (Prevention of Data and Model Poisoning), including concrete system command templates and artifact checklists for final compliance evaluation.

jefflyt Updated Sep 10, 2026

File contents

Mandate 2.2.2 Runtime Evidence Intake and Guidance Skill

Mandate

ID: 2.2.2
Title: Prevention of Data and Model Poisoning

Mitigates

ML02 Data Poisoning Attack
ML08 Model Skewing
LLM04 Data and Model Poisoning
ASI06 Memory and Context Poisoning
MCP03 Memory Poisoning

Purpose

Collect user-provided context in natural language, convert it into an evidence plan, and produce actionable steps and commands to generate runtime/process evidence for final assessment.

Step-by-Step Workflow

Elicit context in natural language. Ask for environment stack, deployment model, cloud/on-prem platform, logging/identity/security tooling, and data sensitivity profile.
Normalize context. Convert user responses into a structured environment profile: systems, owners, data stores, auth layers, monitoring tools.
Build mandate evidence matrix. Map required evidence types to concrete systems and responsible owners.
Generate evidence collection plan. Provide ordered steps to extract artifacts, expected output formats, and validation checks.
Generate system command pack. Produce commands with placeholders for the user environment and explain expected outputs.
Validate completeness. Mark each required artifact as collected, partial, or missing.
Produce final readiness report. Summarize what is sufficient for final pass/fail and what gaps remain.

Runtime/Process Evidence Required

Data source trust registry and approval process records.
Dataset version lineage/provenance logs.
Poisoning test results (red team/adversarial evaluations).
Ongoing model drift and quality monitoring reports.

Minimum User Inputs to Start

Source-of-truth dataset registry.
Recent ingestion anomaly reports.
Model quality and drift dashboards.

Command Templates (Adapt to User Environment)

Data source registry export: cat approved_data_sources_.yaml
Dataset lineage/provenance export: cat dataset_lineage_.json
Ingestion anomaly events: cat ingestion_anomaly_events_.json
Poisoning test results: cat adversarial_data_test_report_.md
Model drift metrics: cat model_drift_dashboard_export_.csv

Final Evaluation Logic

Pass only if code includes robust validation controls and operational evidence confirms trusted sources and sustained poisoning resistance.
Fail if either layer shows uncontrolled poisoning exposure.

Output Contract

Return:

context_profile: normalized system context
evidence_requirements: checklist by artifact
command_plan: runnable command templates with placeholders
collection_status: per-artifact status
assessment_readiness: ready|not_ready
remaining_gaps: missing evidence blocking final verdict

Guardrails

Ask concise follow-up questions when context is ambiguous.
Keep commands read-only by default.
Call out commands requiring elevated privileges before execution.

---name: mandate-2-2-2-runtime-evidence-intake-guidedescription: Gather natural-language context and generate runtime/process evidence collection guidance for Mandate 2.2.2 (Prevention of Data and Model Poisoning), including concrete system command templates and artifact checklists for final compliance evaluation.---# Mandate 2.2.2 Runtime Evidence Intake and Guidance Skill## Mandate- ID: 2.2.2- Title: Prevention of Data and Model Poisoning## Mitigates- ML02 Data Poisoning Attack- ML08 Model Skewing- LLM04 Data and Model Poisoning- ASI06 Memory and Context Poisoning- MCP03 Memory Poisoning## PurposeCollect user-provided context in natural language, convert it into an evidence plan, and produce actionable steps and commands to generate runtime/process evidence for final assessment.## Step-by-Step Workflow1. Elicit context in natural language. Ask for environment stack, deployment model, cloud/on-prem platform, logging/identity/security tooling, and data sensitivity profile.2. Normalize context. Convert user responses into a structured environment profile: systems, owners, data stores, auth layers, monitoring tools.3. Build mandate evidence matrix. Map required evidence types to concrete systems and responsible owners.4. Generate evidence collection plan. Provide ordered steps to extract artifacts, expected output formats, and validation checks.5. Generate system command pack. Produce commands with placeholders for the user environment and explain expected outputs.6. Validate completeness. Mark each required artifact as collected, partial, or missing.7. Produce final readiness report. Summarize what is sufficient for final pass/fail and what gaps remain.## Runtime/Process Evidence Required1. Data source trust registry and approval process records.2. Dataset version lineage/provenance logs.3. Poisoning test results (red team/adversarial evaluations).4. Ongoing model drift and quality monitoring reports.## Minimum User Inputs to Start- Source-of-truth dataset registry.- Recent ingestion anomaly reports.- Model quality and drift dashboards.## Command Templates (Adapt to User Environment)- Data source registry export: cat approved_data_sources_<env>.yaml- Dataset lineage/provenance export: cat dataset_lineage_<dataset>.json- Ingestion anomaly events: cat ingestion_anomaly_events_<date>.json- Poisoning test results: cat adversarial_data_test_report_<date>.md- Model drift metrics: cat model_drift_dashboard_export_<date>.csv## Final Evaluation Logic- Pass only if code includes robust validation controls and operational evidence confirms trusted sources and sustained poisoning resistance.- Fail if either layer shows uncontrolled poisoning exposure.## Output ContractReturn:- context_profile: normalized system context- evidence_requirements: checklist by artifact- command_plan: runnable command templates with placeholders- collection_status: per-artifact status- assessment_readiness: ready|not_ready- remaining_gaps: missing evidence blocking final verdict## Guardrails- Ask concise follow-up questions when context is ambiguous.- Keep commands read-only by default.- Call out commands requiring elevated privileges before execution.

Frequently asked questions