agentleFS
SkillMedevSkillMedev/security-compliance-hardening/skills/soc2-evidence-helper/SKILL.md

soc2-evidence-helper

Skill0 starsChanged 41 days agoMITView on GitHub

Maps engineering controls to SOC 2 Trust Service Criteria, builds a continuous evidence-collection plan with cadences per control family, and produces the control-to-evidence table auditors work from. Use when someone asks "what evidence do we need for SOC 2", "map our controls to the Trust Service Criteria", "how do we prepare for Type II fieldwork", "the auditor asked for access reviews", or is closing findings or standing up a compliance program. Do NOT use for building an incident postmortem - use postmortem-writer instead; do NOT use for finding and prioritizing actual vulnerabilities - use vulnerability-triage instead; this skill organizes proof that controls operate, it does not implement the controls.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When a control has a gap (it was not operating for part of the period), disclose proactively with a documented remediation date.

    line 59
  2. Distinctive:

    Do not collect evidence reactively before fieldwork - Type II samples the entire period.

    line 84
  3. Distinctive:

    Do not fabricate or backdate artifacts.

    line 85
  4. Distinctive:

    Do not leave any control row ownerless - unowned controls are the ones that lapse mid-period.

    line 86
  5. Distinctive:

    Do not hand auditors more than the control requires.

    line 87
  6. Distinctive:

    Do not claim encryption or backups without configuration or restore-test proof.

    line 88
  7. Distinctive:

    Do not skip the annual tabletop - an untested incident response plan is a standard finding.

    line 89
  8. Distinctive:

    Every in-scope criterion has at least one mapped control with artifact, system, owner, and cadence.

    line 93
  9. Distinctive:

    Every cadence meets or beats the table above.

    line 94

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

SOC 2 Evidence Helper

SOC 2 audits test whether security controls exist (Type I, point in time) and operate consistently over a review period (Type II, typically 3-12 months). Engineering teams lose the most time gathering evidence reactively in the month before fieldwork - and for Type II that scramble cannot work, because auditors sample from the whole period and a gap in month two is already a finding. Build collection into normal operations instead.

Operating procedure

Step 1: Gather inputs

  • Type I or Type II, and the review period dates. Default assumption: first audit is Type I, followed by a 6-month Type II.
  • Which Trust Service Criteria are in scope (Security is mandatory; the rest are contract-driven). If unknown, check customer contracts and security questionnaires - scope only what customers demand.
  • The systems of record: IdP (Okta, Entra), source control, CI, cloud provider, ticketing, HR system, vulnerability scanner.
  • Any prior findings or bridge letters. Label unknowns as unknowns.

Step 2: Scope the criteria before collecting anything

Auditors only test criteria in scope:

  • Security (CC - required): logical access, change management, risk assessment, incident response, monitoring.
  • Availability (A): uptime commitments, capacity planning, backup and recovery.
  • Processing Integrity (PI): complete and accurate processing, error handling.
  • Confidentiality (C): data classification, encryption in transit and at rest, NDA controls.
  • Privacy (P): personal data lifecycle - notice, consent, retention, deletion.

Step 3: Build the control-to-evidence map

For each in-scope criterion, name the control, the evidence artifact, the system of record, the owner, and the cadence. This table is the audit's backbone - create it in a shared doc before fieldwork. Never create artifacts that do not reflect how the system actually operates: auditors test consistency, and fabricated evidence is itself a finding.

Practitioner cadences by control family:

Control familyCadenceRed line
Access reviews (CC6.1)Quarterly export + sign-offOffboarding deprovisioned within 24 hours
Change management (CC8.1)Continuous (every merged PR)100% of production changes via reviewed PR + passing CI
Monitoring/alerting (CC7.2)Continuous config + monthly alert-triage sampleEvery sampled alert shows an investigation trail
Vulnerability management (CC7.1)Scans at least monthly; pen test annuallyCriticals remediated within 15 days, highs within 30
Backup/recovery (A1.2)Backups continuous; restore test quarterly at minimum, annually as the floorA restore actually performed, not just configured
Risk assessment (CC3)Annual formal assessmentDocumented with owners and treatment decisions
Incident response (CC7.3-7.5)Plan reviewed annually; at least one tabletop per yearPost-incident reviews exist for real incidents
Policy attestationsAnnual, plus onboardingTimestamped acknowledgment per employee
Security awareness trainingAnnual, tracked per employeeCompletion records with dates

Step 4: Automate collection into a compliance folder

  • Access reviews: scheduled IdP export to a designated, access-controlled evidence folder - not ad hoc screenshots.
  • Vulnerability scans: schedule the scanner and archive full reports automatically; never run scans only when the auditor asks.
  • Pen test: retain the full report including findings and retest results, not just the executive summary.
  • Policy attestations: a GRC tool or a timestamped form; manual sign-off with no timestamp is not auditable.
  • Encryption and infrastructure config: export IaC or config screenshots as evidence - "enabled" without proof of correct configuration is a common finding.

Step 5: Run the auditor interaction deliberately

  • Provide the minimum necessary evidence - do not over-share system diagrams or design docs the control does not require.
  • When a control has a gap (it was not operating for part of the period), disclose proactively with a documented remediation date. Auditors respond better to transparency than to discovering gaps in sampling.
  • For Type II, maintain artifacts continuously across the whole period; auditors sample from all of it.

Worked artifact: evidence-collection table

CONTROL-TO-EVIDENCE MAP - [FILL: company] - Period: [FILL: start] to [FILL: end]

| TSC ref | Control statement                          | Evidence artifact                     | System of record | Cadence     | Owner  | Status |
|---------|--------------------------------------------|---------------------------------------|------------------|-------------|--------|--------|
| CC6.1   | Quarterly user access reviews              | IdP export + reviewer sign-off        | [FILL: Okta]     | Quarterly   | [FILL] | [FILL] |
| CC6.2   | Deprovisioning within 24h of offboarding   | Offboarding tickets w/ timestamps     | [FILL: HRIS+IdP] | Per event   | [FILL] | [FILL] |
| CC8.1   | All prod changes via reviewed PR + CI      | Branch protection config + PR sample  | [FILL: GitHub]   | Continuous  | [FILL] | [FILL] |
| CC7.1   | Monthly vuln scans, criticals fixed <15d   | Scan reports + remediation tickets    | [FILL: scanner]  | Monthly     | [FILL] | [FILL] |
| CC7.2   | Alerting on anomalous activity             | SIEM/CloudTrail config + alert sample | [FILL]           | Monthly     | [FILL] | [FILL] |
| A1.2    | Backups with tested restores               | Backup config + restore-test record   | [FILL]           | Quarterly   | [FILL] | [FILL] |
| CC7.3   | Incident response plan tested via tabletop | Plan doc + tabletop notes             | [FILL]           | Annual      | [FILL] | [FILL] |

Deliverable

Produce a control-to-evidence map covering every in-scope criterion, with artifact, system of record, owner, cadence, and status per row, plus a gap list with remediation dates for any control not yet operating.

Do NOT

  • Do not collect evidence reactively before fieldwork - Type II samples the entire period.
  • Do not fabricate or backdate artifacts; auditors test consistency and fabrication is itself a finding.
  • Do not leave any control row ownerless - unowned controls are the ones that lapse mid-period.
  • Do not hand auditors more than the control requires.
  • Do not claim encryption or backups without configuration or restore-test proof.
  • Do not skip the annual tabletop - an untested incident response plan is a standard finding.

Quality bar

  • Every in-scope criterion has at least one mapped control with artifact, system, owner, and cadence.
  • Every cadence meets or beats the table above; every automatable export is scheduled, not manual.
  • Known gaps are disclosed with remediation dates before the auditor finds them.
  • The four classic engineering findings are pre-empted: 24-hour deprovisioning, documented change management, configuration proof for encryption, and a tested incident response plan.

Escalation

This skill is not legal or compliance advice; the audit opinion belongs to a licensed CPA firm, and scoping questions with contractual consequences belong with counsel. Route incident writeups to postmortem-writer, vulnerability prioritization to vulnerability-triage, and secrets handling to secrets-hygiene.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.