Agent-ready workflow · intelligence

Disclosure to Stack Relevance

Turn a CVE or public report into a stack-specific hypothesis without replaying an exploit blindly.

cvedisclosuresstackreachabilitypatch-analysis
Stable IDworkflow:disclosure-to-stack-relevanceSource record dateHuman reviewNot recordedProvenanceCurated workflow graph

Enter with artifacts, leave with evidence.

Use when
  • new CVE
  • critical disclosure
  • known exploited issue
  • matching product
Prerequisites
  • exact product evidence
  • version or deployment clues
  • current scope
MCP retrieval

Agents can search this workflow, retrieve the complete graph, or request one stage through the read-only Vulns.co MCP connector.

01

Separate source facts

Keep vendor advisory, CVE, CVSS, EPSS, KEV, report narrative, and Vulns.co analysis in distinct evidence lanes.

Inputs
  • primary advisory
  • stack evidence
Outputs
  • source fact table
Evidence gate
Every claim has a named source and date.
Negative control
Popularity or severity alone does not establish reachability.
Stop condition
Do not use third-party exploit code as a fingerprint.
02

Match exact deployment conditions

Compare product, build, module, configuration, protocol, exposure, and authentication prerequisites.

Inputs
  • reachable service evidence
Outputs
  • applicability matrix
Evidence gate
The affected component is reachable and plausibly matches the advisory conditions.
Negative control
Similar branding or shared framework is insufficient.
Stop condition
Stop if product identity or scope is uncertain.
03

Extract the root primitive

Translate the disclosure into parser, authorization, file, request, memory, or state-machine behavior.

Inputs
  • source fact table
Outputs
  • behavior hypothesis
  • safe discriminator
Evidence gate
The discriminator tests the primitive rather than a destructive exploit chain.
Negative control
A known-unaffected or disabled path is identified where possible.
Stop condition
No crash, persistence, or sensitive data access on production.
04

Run a bounded applicability check

Use the cheapest non-destructive observation permitted by policy.

Inputs
  • safe discriminator
Outputs
  • supported, rejected, or unknown status
Evidence gate
The result is reproducible and does not overstate exploitability.
Negative control
The same check against the non-affected path differs as expected.
Stop condition
Escalate only with explicit scope and a separate safe plan.

Continue with the right depth.

References