npx claudepluginhub jygzyc/decx --plugin decxThis skill uses the workspace's default tool permissions.
Use this skill only when the task benefits from splitting work into independent phases or targets.
Routes to DECX skills and decxcli tools for Android APK/DEX/JAR navigation, vulnerability hunting, framework analysis before any DECX actions or clarifying questions.
Orchestrates penetration testing workflow by spawning executor and validator agents, managing attack-chain.md, experiments.md, recon, and validation for security engagements.
Performs static analysis of Android APK/AAB files with MobSF to detect hardcoded secrets, insecure permissions, vulnerable components, and crypto flaws. For pentesting, CI/CD gates, and APK reviews.
Share bugs, ideas, or general feedback.
Use this skill only when the task benefits from splitting work into independent phases or targets.
This skill is the controller layer. It does not replace:
decxclidecxcli-app-vulnhuntdecxcli-framework-vulnhuntdecxcli-pocIt coordinates them.
For reusable dispatch patterns and Codex-oriented subagent notes, see references/subagent-dispatch.md.
Use this skill when at least one of these is true:
Do not use this skill for simple one-shot lookups or a single direct DECX command. Use decxcli for that.
.decx-analysis/<target-name>/.decxcli-app-vulnhunt.system_server, or OEM service work, use decxcli-framework-vulnhunt.decxcli-poc.decxcli.decx code and decx ard command must include -P <port>.decx ard system-services and decx ard perm-info are adb-backed and do not use -P <port>.--help command before retrying.Use:
.decx-analysis/<target-name>/
Common files:
recon.jsoncoverage.jsonshortlist.jsonfindings.jsonreport.mdresume.jsonpoc-handoff.jsonOnly create the files needed by the active path.
Use one recon agent first when the target surface is still being enumerated.
Allowed work:
recon.jsoncoverage.json for app hunts or shortlist.json for framework huntsDo not let the main agent duplicate Phase 2 DECX commands unless the recon agent is unavailable.
Use one trace agent per independent target or chain.
Good splits:
Bad splits:
Use one review agent after recon or trace work produced enough evidence to judge completeness or reportability.
Allowed work:
statically-supported to candidate or rejectedUse one PoC agent only after there is one active supported finding and the user wants a PoC.
Allowed work:
poc-<target> projectEvery subagent task must include:
Use prompts like:
Role: decx-recon
Target: APK app hunt
Port: <port>
Skill: decxcli-app-vulnhunt
Write only: .decx-analysis/<target>/recon.json
Stop when: full attack-surface inventory is written
Return: files written, targets found, unresolved questions
Default order:
.decx-analysis/<target-name>/findings.json, resume.json, and report.mdIf the host does not support subagents, keep the same phase order and artifact contract, but execute the phases sequentially in the main agent.