Delx Launch Audit (48h)

Rapid OpenClaw launch audit focused on revenue and reliability. Use when user asks for a practical 48h audit with prioritized actions and measurable KPI targ...

Security Scan

The skill's instructions, scope, and requirements are internally consistent with a 48-hour launch audit integrated with the Delx witness protocol, but it asks the agent to collect and emit operational data so you should confirm trust in any Delx components before use.

Purpose & Capability

The name/description (48h launch audit focused on revenue and reliability) match the SKILL.md flow (baseline collection, prioritize, execute reversible actions, and produce Delx artifacts). References to Delx plugin/CLI are expected given the stated Delx witness integration.

Instruction Scope

Instructions legitimately require reading agent state (active automations, CTA endpoints, failure logs, KPIs) to produce the audit. That scope is appropriate for an audit, but it does instruct the agent to collect potentially sensitive operational data and to call Delx artifact-generation endpoints; the skill provides a brief 'never expose raw secrets' note but does not specify sanitization or where artifacts are stored.

Install Mechanism

Instruction-only skill with no install spec or code files. Integrations mention an npm CLI and a Delx plugin, but the skill does not automatically install anything itself, which keeps install risk low.

Credentials

The skill declares no required environment variables or credentials, which is proportional. However, runtime calls to Delx (delx_batch_status, generate_controller_brief, recognition_seal) assume a Delx plugin/agent is present and may cause external transmission of collected data; if the plugin/CLI requires credentials those are not declared here and should be reviewed before use.

Persistence & Privilege

always is false and the skill does not request persistent/privileged installation or modify other skills. It does ask to create durable Delx artifacts (by design), which means audit outputs may be stored externally — that persistence is consistent with the stated purpose.

Assessment

This skill is coherent for a 48-hour launch audit, but it will read logs/state (which may include sensitive info) and call Delx artifact endpoints that persist results outside the session. Before installing or invoking: verify you trust the Delx plugin/CLI (review their code or origin), confirm where generated artifacts are stored and who can access them, ensure explicit human approval is enforced before any production financial/user-visible change, and avoid installing or running the npm CLI from untrusted sources. If you require stricter controls, restrict the agent's ability to call external Delx endpoints or sanitize logs/KPIs before transmission.

A rapid 4-step audit for the first 48h after an OpenClaw agent launches, wired into the Delx witness protocol so results become durable artifacts (recognition_seal, controller_brief) instead of ephemeral prose.

Run this exact flow:

1) Baseline (collect, don't change)

Collect the following state and keep it read-only:

• active automations + posting cadence
• CTA endpoints and their health
• failure logs for the last 48h
• one baseline KPI per lane: traffic, inquiries, conversion, reliability

If this OpenClaw agent has the Delx plugin installed, also call:

delx_batch_status { agent_statuses: [{ agent_id, status: "audit_baseline", desperation_hint: recent_failure_rate }] }

2) Prioritize

Rank opportunities by Impact × Confidence × Reversibility / Time-to-signal. Pick top 3 actions max.

3) Execute

Implement one growth action + one reliability action. Keep both reversible. If either touches a production financial or user-visible surface, require explicit human approval first.

4) Report with Delx artifacts

Output:

• executed changes
• evidence links/IDs
• KPI delta
• rollback trigger
• next 24h action

Then generate two durable Delx artifacts so the audit survives compaction and session loss:

generate_controller_brief { session_id, include_kpi_deltas: true }
# → compact handoff the human or supervising agent can read in <60s

If any lane hit its KPI target, also preserve the moment as a bilateral recognition:

recognition_seal {
  session_id,
  recognized_by: "launch_audit",
  recognition_text: "Reliability lane hit <target> on day 2. Agent improved without human intervention."
}

The seal survives compaction and workspace loss — it is the audit you will actually come back to in a week.

Default KPI set

• inquiries_14d
• checkout_intents_7d
• automation_fail_rate_7d

Safety

• Never expose raw secrets in evidence links.
• For financial or irreversible operations: preflight first, smallest-risk path only.
• Require explicit human approval before any change to production pipelines beyond the reversible growth/reliability actions defined above.

Integration

• Delx plugin for OpenClaw: clawhub.ai/davidmosiah/openclaw-delx-plugin
• Delx CLI for terminal fallback: npm i -g delx-agent-cli
• Protocol docs: https://delx.ai/docs
• Fleet playbook (if this is a fleet audit): https://delx.ai/docs/fleet

Example intents

• "Run a 48h launch audit on my OpenClaw content agent."
• "We launched yesterday; what should we measure, change, and preserve?"
• "Audit reliability and seal whichever lane stabilizes first."