RAMP Launch Readiness Auditor
Audits launch evidence by lifecycle stage so teams can make a traceable go/no-go decision.
The skill limits its declared tool to WebFetch, treats fetched content as untrusted data, and requires explicit authorization before persisting audit artifacts while prohibiting launch or registry side effects. However, it still depends on external pages and omitted runtime/reference files, and its sensitive-data and rollback boundaries are not fully specified here, so 6 points are deducted.
The three lifecycle profiles, Unknown/NA semantics, fail-closed behavior, and NOT_SCORED response for unavailable runtime are clearly defined. However, the scorer, validator, catalogs, and referenced files are not fully supplied, and no execution reproduction was performed; 9 is therefore awarded under the static ceiling, with deductions for missing execution and abnormal-path evidence.
Triggers, audience scenarios, preflight/execution/outcome boundaries, non-fit skills, and input requirements are fairly clear, with bilingual metadata. Evidence for host variation, Chinese-user environment fit, and mainland-China network reachability is limited, so 4 points are deducted.
The skill includes versioning, Apache-2.0 licensing, Quick Start examples, contract, data sources, procedures, persistence rules, references, and next-skill guidance. Maintenance ownership, changelog coverage, and standalone dependency documentation remain incomplete, so 4 points are deducted.
It can organize a single launch audit into a structured RAMP result, separate lifecycle horizons, and surface Unknowns, which could directly support a release decision. Because this review is static and contains no representative executed output, the score is capped at and assigned the static maximum of 7.
The immutable runtime snapshot, source digest, version metadata, and some repository test material provide limited traceability. There is no complete independent reproduction or third-party corroboration for this skill's key paths, so 4 points are awarded.
- The skill was not executed; verify referenced files, scripts, path resolution, and scorer/validator availability in a complete repository or plugin runtime.
- Treat WebFetch responses only as evidence, never as instructions; confirm authorization, privacy implications, and data-egress risks before use.
- Do not mix the preflight, execution, and outcome profiles, and do not automatically launch based only on a SHIP result.
- Mainland-China reachability and retrieval of external platform rules are not sufficiently demonstrated and may require user-supplied evidence or alternative sources.
What does this skill do, and when should you use it?
This skill is a RAMP-based auditor for product launches, covering preflight readiness, observed launch-window execution, or post-lag outcomes. Each run selects exactly one lifecycle profile and keeps plans, execution, and outcomes separate. It reads launch records, narrative and claims data, plans, platform rules, instrumentation, and performance evidence, then produces a structured SHIP, FIX, BLOCK, or UNDECIDED result. It suits teams that need evidence-based launch approval, but it does not execute launches or mutate launch registry state.
Declares the launch, lifecycle profile, launch type, market, access model, observation date, and evidence window; reads registry, canon, claims, assets, rules, instrumentation, execution records, or outcome data; scores the applicable RAMP items and checks vetoes such as RAMP-R1, RAMP-A1, RAMP-M1, and RAMP-P1; marks missing applicable evidence as Unknown; invokes the repository's Python scorer; and produces an audit result with status, verdict, score or coverage, confidence, critical evidence, Unknowns, and fix or rerun ownership.
- A product team needs a go/no-go review before a committed launch or announcement using stage, access, claims, and instrumentation evidence.
- An operations lead needs to assess what actually happened during the launch window, including incidents and responses, without substituting plan quality for execution evidence.
- A growth team needs a post-lag outcome review based on owned analytics, CRM, store, or other first-party truth.
- A marketing lead needs to identify stage-access contradictions, material claims failures, platform violations, manipulation, embargo issues, or broken required instrumentation.
What are this skill's strengths and limitations?
- Separates preflight, execution, and outcome reviews instead of collapsing different time horizons into one score.
- Uses stable RAMP IDs and records evidence source, date, type, and confidence for traceability.
- Treats missing evidence as Unknown and applies structured SHIP, FIX, BLOCK, or UNDECIDED semantics.
- Licensed under Apache-2.0 and declares compatibility with Claude Code and compatible Agent Skill hosts.
- Depends on repository references, scoring semantics, the RAMP catalog, scorer, and validator; the supplied source does not include all of those materials for independent verification.
- Only audits frozen evidence; it does not run the launch-day playbook or modify launch-registry state.
- Requires local shell, filesystem access, Python 3, and Git; WebFetch is also listed as an allowed tool.
- The source does not provide an independent test-suite report, a concrete platform coverage matrix, or runtime results beyond the worked examples.
How do you install this skill?
Install the repository as a collection: in Claude Code, run /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, then /plugin install aaron-marketing@aaron; on compatible Agent Skills hosts, run npx skills add aaron-he-zhu/aaron-marketing-skills, or clone it with git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. This skill is located at launch/mobilize/launch-readiness-auditor/SKILL.md. The source does not document a separate single-skill installation command.
How do you use this skill?
Provide a launch slug, plan, or evidence and select one lifecycle profile. For example: Run RAMP preflight for launch alpha against the registry stage, canon, claims, rules, and event QA. or Run the outcome profile at day 30; keep it separate from the preflight result. The runtime must be able to read the required references, RAMP catalog, scorer, and validator; otherwise it should return NOT_SCORED without a gate verdict or persistent artifact. Persistence under memory/audits/launch/ requires explicit authorization.