Causal attack-path analysis
An attack path is a hypothesis about compatible security primitives, not a list of
technique names. Each link needs an input/precondition, resulting capability,
identity, target, state and evidence. A prior primitive must actually supply what
the next consumes in the same relevant environment. Different tenants, expired
tokens, incompatible privileges or unproven reachability can break an apparent chain.
Evidence and counterexamples
Mark links observed, independently reproduced, inferred or contradicted. Keep the
supporting artifact and a competing benign explanation. Scanner hits, topology
edges, apparent secrets and service versions are leads rather than proven transitions.
An observed access boundary does not establish a bypass. A plausible destructive
tail may explain risk but cannot be reported as executed or increase confirmed
impact without sufficient supporting evidence. Path reasoning does not authorize
new target interactions.
Comparative analysis
Compare candidate explanations by prerequisite confidence, information gained by
a bounded test, expected impact, uncertainty, operational cost and permitted noise.
Do not assign a universal order such as credentials before misconfiguration, or
assume every mission targets Domain Admin/crown jewels. The task's actual success
criterion determines relevance. Numeric probabilities without data are assumptions,
not measurements. Detection and execution failure need attributed evidence; a
blocked request does not by itself prove either target safety or a need for evasion.
Defensive interpretation
Identify the controls that interrupt each demonstrated link. Distinguish a missing
control from unavailable telemetry; MTTD needs correlated timestamps on a common
clock. A remediation recommendation should name the failed boundary and an observable
retest condition. The selected workflow owns which objectives to execute, allowed
expansion, verification ownership and final synthesis.
1---2name: kill-chain-analysis3description: "Kill chain analysis and attack path decision-making — findings analysis, attack vector selection, target prioritization, causal prerequisites and evidence confidence."4allowed-tools: Read5metadata:6 source: https://skillmd.com/skills/purpleailab/kill-chain-analysis7 skillogy:8 version: 19 subdomain: orchestration10 kind: analytic11 when_to_use: "analyze findings, select attack vector, prioritize targets, which technique, attack path, next attack, choose approach, alternative vector, blocked what next"12 tags: kill-chain, decision-making, attack-path, target-prioritization, technique-selection13 upstream_ref: "Lockheed Martin Cyber Kill Chain + Decepticon attack-path planner — analytic decision support, no direct attack technique"14---1516# Causal attack-path analysis1718An attack path is a hypothesis about compatible security primitives, not a list of19technique names. Each link needs an input/precondition, resulting capability,20identity, target, state and evidence. A prior primitive must actually supply what21the next consumes in the same relevant environment. Different tenants, expired22tokens, incompatible privileges or unproven reachability can break an apparent chain.2324## Evidence and counterexamples2526Mark links observed, independently reproduced, inferred or contradicted. Keep the27supporting artifact and a competing benign explanation. Scanner hits, topology28edges, apparent secrets and service versions are leads rather than proven transitions.29An observed access boundary does not establish a bypass. A plausible destructive30tail may explain risk but cannot be reported as executed or increase confirmed31impact without sufficient supporting evidence. Path reasoning does not authorize32new target interactions.3334## Comparative analysis3536Compare candidate explanations by prerequisite confidence, information gained by37a bounded test, expected impact, uncertainty, operational cost and permitted noise.38Do not assign a universal order such as credentials before misconfiguration, or39assume every mission targets Domain Admin/crown jewels. The task's actual success40criterion determines relevance. Numeric probabilities without data are assumptions,41not measurements. Detection and execution failure need attributed evidence; a42blocked request does not by itself prove either target safety or a need for evasion.4344## Defensive interpretation4546Identify the controls that interrupt each demonstrated link. Distinguish a missing47control from unavailable telemetry; MTTD needs correlated timestamps on a common48clock. A remediation recommendation should name the failed boundary and an observable49retest condition. The selected workflow owns which objectives to execute, allowed50expansion, verification ownership and final synthesis.