Proof Point Packager
Turn approved evidence into reusable proof assets placed beside the claims they support.
The skill limits itself to packaging ledger-approved proof, treats supplied material as untrusted data, forbids fabrication, and uses authorized proposal events plus confirmation before memory writes. Data sources, write locations, and external effects are relatively transparent. Deducted 5 points because the supplied static material does not prove that the host enforces authorization, rollback, or isolation, and it lacks a concrete sensitive-data minimization procedure.
Triggers, NEEDS_INPUT behavior, required module fields, self-checks, and termination conditions are fairly explicit, with diagnosable missing-input paths. Deducted 12 points because the skill depends on upstream skills, project memory, and registry-events.py; its key path was not reproduced statically and no skill-specific tests or abnormal-input coverage were provided. Static calibration therefore keeps this at or below 10.
Audience, use cases, non-fit boundaries, inputs, and outputs are clearly described, with English and Chinese terminology. Tier-1 operation and user-provided data improve environment fit, and the core function does not depend on overseas services. Deducted 4 points because it requires an existing message house, claims ledger, story bank, and authorization workflow; independent-use boundaries and host-specific reachability are not evidenced.
The file has frontmatter, versioning, Apache-2.0 licensing, Quick Start, a contract, data-source notes, procedures, handoff guidance, limitations, and next-skill routing. Deducted 4 points because the supplied material lacks skill-specific example outputs, FAQs, a changelog, an explicit maintainer, or a clear update path; publisher identity is also unverified.
The expected artifacts, required fields, proof-gap list, and handoff target are concrete and could be directly useful when prerequisites exist. Deducted 9 points because no representative generated output or execution evidence is supplied, and the skill cannot complete independently without prerequisite state. Static calibration therefore keeps this at or below 7.
Claim IDs, source labels, as-of dates, permission boundaries, and gap routing create an auditable structure; repository-level CI and test materials provide some supporting evidence. Deducted 6 points because the supplied tests do not cover this skill's key behavior and there is no third-party execution evidence or independent multi-source corroboration.
- Before use, confirm each claims-ledger claim ID, approval status, evidence rights, and as-of date; this skill does not adjudicate claims for you.
- Memory writes and registry proposals create persistent state changes, so use the host authorization and confirmation flow and verify the actual registry-events.py deployment.
- This assessment is static only; output formatting, cross-skill references, and abnormal-input handling were not executed or independently verified.
What does this skill do, and when should you use it?
Proof Point Packager is a Land-phase brand-narrative skill that converts claims-ledger-approved evidence into stat cards, case snippets, testimonial blocks, and comparison proofs. Every module is pinned to a message-house pillar and the ledger claim ID it supports, with a Measured or User-provided label and as-of date. It flags unsupported pillars but does not adjudicate claims, invent benchmarks, or score narrative quality.
Reads message-house pillars, approved claims-ledger entries, story-bank units, and user-provided case data, benchmark exports, or permitted quotes; writes the proof module set to memory/narrative/proof-point-packager/; tags each module with its pillar, claim ID, source label, and as-of date; routes missing or not-yet-ledgered proof as [needs source] candidate proposals through registry-events.py; and produces a proof-gap list plus handoff summary.
- A brand team has an approved message house and needs reusable stat cards for each pillar.
- A marketer has customer case material and permitted quotes that need to become claim-adjacent narrative assets.
- A content team needs to identify message pillars that lack approved proof.
- A narrative team is preparing for TALE review and needs evidence assets tied to specific claims.
What are this skill's strengths and limitations?
- Has a narrow, explicit scope focused on packaging approved proof.
- Produces concrete asset types: stat cards, case snippets, testimonial blocks, and comparison proofs.
- Requires pillar, ledger claim ID, source label, and as-of date on each module.
- Reports proof gaps instead of fabricating numbers or benchmarks.
- Requires an existing message house, story bank, and approved claims ledger.
- Does not resolve missing or disputed evidence.
- Saving results to durable memory requires user confirmation.
- Evidence rights, factual validity, and claim approval remain dependent on upstream processes.
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 an Agent Skills-compatible host 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. The source does not document a proof-point-packager-specific single-skill install command.
How do you use this skill?
Provide an existing message house, story bank, and approved claims ledger, then use: Package proof points for [product] from the approved claims ledger. Pillars: [list or all three]. Another supported prompt is: Build reusable stat cards and case snippets for each message-house pillar, each pinned to its claim ID. If no message house exists, the skill should stop with NEEDS_INPUT.
How does this skill compare with similar options?
Unlike offer-claims-registry, this skill packages approved proof rather than adjudicating or substantiating claims. Unlike narrative-cascade-planner, it does not map modules onto specific channel surfaces. Unlike narrative-quality-auditor, it does not calculate TALE results or score E/L.