菜单
Systematic web search for news, media, policy, and industry coverage of FORTIS Lab publications, tools, and research. Use when the user asks for a news audit, media coverage check, broader impact evidence, or visibility search for their work.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/yzhao062/yzhao062.github.io --skill news-search命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Add or update content that must appear across the website, the LaTeX CV, and the GitHub profile README. Use when the user mentions a new paper, award, grant, service role, teaching course, PhD student, open-source project, research direction, or headline metric. Use it for any other content that overlaps those surfaces.
Decide or validate a research-program name, site headline, lab tagline, or talk framing by measuring where the person actually surfaces, instead of arguing about which phrase sounds better. Use when choosing an umbrella term, renaming a program, or judging whether a phrase is dated or contested. Also use it to check who already owns a phrase, and to re-audit on a schedule whether a public self-description still matches the work.
Systematic bibliometric audit of citing-paper author affiliations across OpenAlex and Dimensions Analytics. Use when the user wants to find which notable institutions (government agencies, foundation model companies, national labs, Big Tech, finance, pharma) have authors who cite their published work. Pair with the news-search skill for full external-impact evidence; this skill is the bibliometric half, news-search is the editorial-coverage half. Also use for legacy prompts: "citation audit", "affiliation audit", "citation-affiliation audit", or "Dimension 9 audit" (this skill was previously Dimension 9 inside news-search; split out 2026-05).
正在显示 SKILL.md
| name | news-search |
| description | Systematic web search for news, media, policy, and industry coverage of FORTIS Lab publications, tools, and research. Use when the user asks for a news audit, media coverage check, broader impact evidence, or visibility search for their work. |
Systematically search for external coverage of Yue Zhao / FORTIS Lab work across all outlet types: tech press, security press, business press, government/policy, industry analysts, science press, AI newsletters, and universities.
Read these files before starting:
data/publications.json — all papers (titles, venues, years, links)data/open-source.json — all tools/libraries (names, stars, URLs)news-coverage-audit.md — previous audit results (skip known items, update stale entries)If parallel workers and web search are available, run dimensions in parallel; otherwise process dimensions sequentially. Batch queries conservatively to stay within tool rate limits. Read these reference files before starting:
references/search-queries.md: query bank (not exhaustive; see triage rules below)references/outlet-registry.md: outlet classification and site: domain listsreferences/search-strategy.md: techniques for finding indirect coverage, when to persist vs. stop, name disambiguation, Cloudflare/SSR-shell fetch tactics, and the Phase B "snippet alone is not verified" rulereferences/candidate-schema.md: Phase A and Phase B candidate record contractreferences/disclaimer-patterns.md: AI-generated, aggregator, translation, templated-database, and blocked-page detectionreferences/domain-registry.md: seed domains by source class and outlet_class values for Phase Areferences/disambiguation-registry.md: cumulative tool-name and person-name collision rules + verified-negative leads from prior rounds (consult before counting any borderline match)scripts/pdf_term_scan.py: PyMuPDF-based FORTIS-term scanner with built-in false-positive filters; canonical Phase B PDF-deep-search tool. Run as python skills/news-search/scripts/pdf_term_scan.py <pdf_path>.Run the two-phase pipeline described below. Phase A gathers candidates into news-search-candidates.jsonl and does not edit news-coverage-audit.md. Phase B verifies each candidate, classifies the survivors, records dropped candidates in the candidate file, and writes only kept rows to news-coverage-audit.md using the tier structure in the Output section. If news-coverage-audit.md does not yet exist, Phase B creates it with the full tier structure and negative-results table as a fresh audit.
The query bank in references/search-queries.md is curated, not exhaustive. It covers high-adoption tools and papers with known media hooks.
Full audit mode: search EVERY paper and tool. Read data/publications.json and data/open-source.json and generate at least one smart-keyword search (Dimension 5) for every single entry. Do not skip any paper or tool. Use distinctive claims or method names, not exact titles. This is critical because coverage can appear for any paper, not just high-profile ones (e.g., COPOD has a dedicated book chapter, GLIP-OOD has a tech blog feature, the computing resources paper influenced CVPR policy).
For items not in the query bank, generate queries at runtime:
Each audit runs in two phases. Phase A discovers candidates without classifying them. Phase B reads each candidate, runs verification checks, and assigns a tier. Splitting the two lets the candidate list be reviewed before any classification work commits to a row in the audit ledger.
Run news-search dimensions D1-D8 plus D10 and emit candidates as JSON-Lines records to news-search-candidates.jsonl at the project root. Handle citation-affiliation evidence through the freshness-gated [[citation-audit]] hook (see "Cross-skill: citation-audit integration" below), not as a Phase A candidate dimension. Do not assign tiers yet. Do not write to news-coverage-audit.md yet.
Each candidate carries the schema in references/candidate-schema.md: URL, title, snippet, surfacing query, outlet class, fetch timestamp, plus empty placeholders for the Phase B fields (flags, direct-mention, tier, notes).
When Phase A completes, present the candidate count grouped by dimension and outlet class to the user. The user (or a reviewer such as Codex) can scan the candidate list and flag wrong query routing or wrong outlet-class tagging before Phase B starts. This is the cheap, parallelizable stage; treat it as re-runnable.
Add news-search-candidates.jsonl to .git/info/exclude (local, untracked) before the first run so git add -A does not stage scratch output.
For each candidate in news-search-candidates.jsonl, fetch the page and apply five checks in order:
Pre-tier filter: first-party / already-tracked / disambiguation drops. Before running the citation rule, drop the candidate if it falls into any of these patterns (each was stepped into during the 2026-05-07 round):
references/disambiguation-registry.md.Direct-mention / topic-validation routing (the citation verification rule in the Output section). If the page names the work, person, lab, co-author, institution, or direct URL per one of clauses 1 to 6, fill direct_mention and continue as coverage. If it does not pass direct mention but clearly covers the same topic area, set tier: "topic-validation" and keep it for the Topic Validation appendix, not a coverage ledger. If it is neither direct coverage nor topic validation, set tier: "dropped" and record the drop reason in notes.
Snippet alone is not verified evidence for Tier 0 / Tier 1 candidates. WebSearch summaries can synthesize content that does not appear in the source (the 2026-05-07 round caught this with GAO-26-108695: snippet claimed TrustLLM citation; manual PDF extraction confirmed the PDF says nothing of the sort). Tier 0 / Tier 1 promotion requires direct fetch of the source — pdf_term_scan.py for PDFs, real-UA HTTP for web pages. If the source is gated and cannot be re-fetched, set tier_guess: phase_b_priority and leave as a candidate; do not count.
Disclaimer / aggregator detection (references/disclaimer-patterns.md). Run the regex sweep on fetched content. Set entries in the candidate's flags[] field. Hard caps:
ai_generated and aggregator are capped at Tier 3 regardless of outlet domain.machine_translated is capped at Tier 3 unless is also set.Phase B writes direct-coverage rows (Tier 0 through Tier 5) to the coverage ledgers, topic-only rows (tier == 'topic-validation') to the Topic Validation appendix, and keeps dropped rows (tier == 'dropped') in news-search-candidates.jsonl for auditability. The full candidates file stays at the project root through the audit so a reviewer can audit drop decisions, not only the kept rows.
references/domain-registry.md lists known high-value source classes (gov / policy PDFs, EU research projects, patents, China tech media, security research blogs, AI-newsletter aggregators, and others) with seed domains. Phase A queries fan out to seeded domains in addition to the open dragnet, never instead of it. The registry is a recall floor, not a filter.
After each audit, harvest the domains of every confirmed Phase B hit and append new ones to the registry under the appropriate class. If no class fits, create one (lowercase-hyphenated name). This is the only way the registry stays current as new outlet types appear; without it, the registry freezes and re-discovery cost recurs.
Once a quarter, run a registry-disabled pass (open dragnet only) to surface new outlet classes the registry has not seen yet. This is what catches the next surprise category.
This lane runs in every run mode, including Quick check and Targeted, and it runs in addition to whatever dimensions that mode selects. It never replaces a dimension. If a run is so narrow that only one lane fits, this is the lane. A round that finds nothing here must say so explicitly in the output rather than omit the lane, so the absence stays visible across rounds.
Why it is mandatory. The agent-auditability line is the lab's current research frontier and the axis the public positioning rests on, so coverage there moves fastest and matters most to catch early. It is also the material the dimension sweep is worst at finding. The artifacts are new, several live under co-author GitHub accounts rather than yzhao062, and much of the relevant writing names a concept such as "agent auditability" rather than a tool name that a keyword sweep would match.
| Work | Identifier |
|---|---|
| Auditable Agents (position paper: five auditability dimensions, three mechanism classes) | arXiv:2604.05485 |
| Aegis (pre-execution firewall for tool calls) | arXiv:2603.12621, github.com/Justin0504/Aegis |
| agent-audit (static security and over-privilege scanner) | arXiv:2603.22853, github.com/HeadyZhang/agent-audit |
| Implicit Execution Tracing (post-hoc attribution after logs are stripped) | arXiv:2603.17445 |
| GRADE (typed two-layer execution and dependency graph) | arXiv:2606.22741, github.com/yzhao062/grade |
| FORTIS (over-privilege benchmark for agent skills) | arXiv:2605.09163 |
| WeClawArena (auditable sandbox and benchmark for cross-user agent collaboration and security) | arXiv:2608.03499 |
auditable (open-source system of record for agent decisions) | github.com/yzhao062/auditable, pypi.org/project/auditable |
| awesome-auditable-ai (curated knowledge base) | github.com/yzhao062/awesome-auditable-ai |
| Auditability Card (machine-checkable disclosure artifact) | named inside Auditable Agents |
Keep this table current. When a new work ships on this line, add it here in the same commit that adds it to data/publications.json or data/open-source.json.
Coverage on this line often names the idea rather than the artifact, so search the vocabulary alongside the tool names: "auditable AI systems", "agent auditability", "accountability layer" for AI agents, "system of record" for agent decisions, decision replay against live state, rollback and compensation for agent actions, tamper-evident agent audit trails, pre-execution mediation, and post-run attribution in multi-agent systems.
Apply the citation verification rule without softening it. Most vocabulary matches will be other people writing about the same idea, which is topic-validation. This lane earns its place by catching the minority that name the work.
aas-1.org), EU AI Act Article 12 logging and traceability, NIST work on agent security, OWASP agentic releases, and any ISO/IEC JTC1 SC42 agent-audit item. These are the surfaces where a term turns into a requirement.Auditable AI is already in descriptive use by FICO, Fraunhofer HHI, IBM, Rezolve AI, Numbers Protocol, and an unrelated company at auditableai.com. AuditBench is Anthropic's alignment-auditing benchmark and is a separate project. FORTIS is the lab name, the benchmark name, and also a bank, a utility, and a hospital chain. Check these before counting any match and record new collisions in references/disambiguation-registry.md.auditable.run is the PI's own company site, so it is first-party and never counts as coverage. External writing about the company does count.Find coverage that names the PI or lab, regardless of which paper or tool.
Search for: name + university + research area + various contexts (news, interview, podcast, keynote, expert quote, award, fellowship, grant). Also search for lab name and industry partner names (Amazon, NVIDIA, Google, Meta, Anthropic, NSF).
See references/search-queries.md § Dimension 1 for the full query list.
Major tools (PyOD, TrustLLM, agent-audit, Aegis, ADBench) may appear in industry deployments, government reports, textbooks, or enterprise case studies without naming the PI.
Search for: each tool name + context keywords (enterprise, deployment, production, fraud detection, cybersecurity, government, NIST, federal, textbook, course, patent, Walmart, NASA, Tesla).
See references/search-queries.md § Dimension 2 for the full query list.
Systematically check each outlet category using site: filters. This is the most important dimension for finding coverage the other dimensions miss.
Categories: security press, business press, top tech press, AI newsletters, science press, government/policy, industry analysts, university/institutional press, developer community.
Generation rule: references/search-queries.md § Dimension 3 provides base queries for the highest-priority outlets. For any outlet domain listed in references/outlet-registry.md that does not have an explicit query in the query bank, generate one at runtime using this template: site:{domain} "anomaly detection" OR "AI auditing" OR "AI agent security" OR PyOD OR TrustLLM. This ensures every registered outlet is checked without requiring the query bank to enumerate all 70+ domains.
Search for the broader trending topic and check if the work appears within coverage. This catches indirect coverage where the paper is relevant but not cited by name.
Examples: ChatGPT geolocation trend (connects to DoxBench), OpenClaw security crisis (connects to agent-audit), OWASP agentic AI landscape (connects to agent-audit/Aegis), anomaly detection open-source landscape (connects to PyOD).
See references/search-queries.md § Dimension 4 for the full query list.
For papers where exact title search fails (most papers), use distinctive result keywords or striking claims from the paper instead.
Examples: "surpassed human performance 95.33% CLADDER" for the causal reasoning paper, "defense training breaks LLM agents 47-77% benign task failure" for The Autonomy Tax.
See references/search-queries.md § Dimension 5 for the full mapping table.
Track high-level citation metrics, appearances in high-impact journals (Nature, Science), enterprise adoption evidence, and downstream tools built on the work.
See references/search-queries.md § Dimension 6 for the full query list.
Search the surfaces where widely-adopted tools spread beyond academic papers and news: education platforms, code ecosystems, non-English press, and developer communities.
Education & courses: Kaggle notebooks, Google Colab examples, Coursera/edX/Udemy course materials, university syllabi, YouTube/Bilibili tutorials, O'Reilly/Manning learning paths Code ecosystem: GitHub code dependents (repos that import PyOD), PyPI/conda-forge download stats pages, Papers with Code tool listings, Hugging Face Spaces built on your tools Dissertations & theses: ProQuest, Google Scholar thesis search, university repository searches Non-English coverage: Chinese tech press (InfoQ CN, CSDN, Zhihu, WeChat public accounts), Japanese (Qiita, Zenn), Korean (Tistory, Velog), European (Heise, Le Monde Informatique, etc.)
See references/search-queries.md § Dimension 7 Education/Ecosystem for queries.
This is critical and cannot be skipped. Web search does not index the text inside PDFs from government reports, congressional testimony, think tank whitepapers, and industry reports. Citations of your work inside these documents are the highest-impact coverage (Tier 0) and are routinely missed by Dimensions 1-6.
U.S. Government (highest priority):
International Government (high priority):
Think tanks & policy institutes:
Foundation model companies (Tier 0 if they cite your work):
These companies publish system cards and safety evaluations as PDFs or long-form web pages that reference academic benchmarks (TrustLLM, HELM, etc.) and tools. They also publish blog posts with embedded citations. Search both the HTML pages and any linked PDFs.
Standards bodies:
Industry whitepapers & analyst reports (Tier 0 if they cite your work by name):
Government PDFs are hosted as static files (e.g., .senate.gov/wp-content/uploads/...pdf). Web search engines index the hosting page but not the text inside the PDF. A search for site:senate.gov TrustLLM returns nothing because the word "TrustLLM" only appears inside the PDF, not on any HTML page. The only way to find these is to identify candidate documents by topic, then search inside the PDFs directly.
Write all results to news-coverage-audit.md at the project root.
An item only counts as coverage if the article names or cites at least one of:
Read the article or its snippet to verify before including it. Items attributed to co-authors or institutions should note this in the entry (e.g., "names: first author X, USC affiliation — Yue Zhao is co-author").
Items that only cover the same topic your work addresses (e.g., "AI agent security is important" without naming your tools) are not coverage. These may be useful as context for grant narratives but must be placed in a separate "Topic Validation (Not Direct Coverage)" appendix, clearly marked as not naming your work.
All tiers below require the citation verification rule above. If a result does not name or cite your work, it does not belong in any tier.
| Tier | Definition | Examples |
|---|---|---|
| Tier 0 | (a) Government reports (U.S. or international: legislative, executive, federal/national agency), international body reports (OECD, UN, EU), or official standards documents; (b) Technical reports, system cards, safety reports, or model cards from major foundation model companies (OpenAI, Anthropic, Google DeepMind, Meta AI, Mistral, xAI, Cohere, etc.); (c) Major consulting/analyst firm reports (McKinsey, Gartner, Forrester, Deloitte, etc.) — all that cite your work by name | U.S. Senate report citing TrustLLM, OpenAI system card citing TrustLLM, Anthropic safety report citing anomaly detection benchmark, Gartner report citing PyOD |
| Tier 1 | Mainstream tech/business/security press, major research institutions (national labs, Hoover, Microsoft Research), or high-impact policy reports that name your work | FLI AI Safety Index using TrustLLM, LLNL article naming TrustLLM, Nature Biotechnology citing DrugAgent |
| Tier 2 | Industry press, institutional PR, or dedicated features that name your work | "DrugAgent" in MarkTechPost, "Yue Zhao" in USC Viterbi News, Databricks blog naming PyOD |
| Tier 3 | Dedicated blog posts, tutorials, or platform integrations naming your tool | KDnuggets PyOD tutorial, Databricks Kakapo built on PyOD, DEV Community Aegis post |
| Tier 4 | Awards, recognitions, encyclopedia entries | Amazon RA, NVIDIA Grant, Grokipedia entry |
| Tier 5 | Academic community only (Hugging Face, alphaXiv, etc.) | Paper pages, GitHub stars, Moonlight reviews |
Separate appendix (not a tier):
A first-party foundation-model-company job posting that names a FORTIS tool as expected operational tooling (e.g., the OpenAI "Technical Intelligence Analyst" Qualifications block naming PyOD as anomaly-detection tooling) qualifies as Tier 0(b)-equivalent only when all of the following hold:
openai.com/careers/..., anthropic.com/jobs/..., deepmind.google/careers/..., ai.meta.com/careers/..., etc.), not a Greenhouse / Lever / Ashby / DFJ Growth / Glassdoor / LinkedIn / Indeed mirror. ATS mirrors are kept under mirrors[] in the candidate record but never count as the load-bearing citation.news-snapshots/<slug>-<YYYY-MM-DD>.{html,pdf}) is in the repo, with a Markdown index file documenting the live URL, capture date, verification method, and verbatim quote. Sidecars must be captured from a logged-in browser session when the live URL is behind Cloudflare; PDF must be re-verified with python skills/news-search/scripts/pdf_term_scan.py <pdf_path>.When all three hold, the candidate goes into Ledger 1 (Government/Policy citations) under Tier 0(b) with a Source URLs row that includes the live URL, mirror URLs, and the snapshot index path. The #8g precedent (news-snapshots/openai-careers-technical-intelligence-analyst-2026-05-07.md) is the reference shape; new entries follow that index format.
If the live URL is reachable but no snapshot exists yet, set tier_guess: phase_b_priority and status: paywall_or_blocked (or candidate with a snapshot-pending note in notes). Do not promote to Ledger 1 from a snippet alone — careers pages go stale within weeks of the role being filled, so an unsnapshotted Tier 0(b) claim becomes unverifiable as soon as OpenAI / Anthropic rotates the URL.
Non-FM-co careers pages (Wells Fargo, Capital One, JPMC, Pfizer, Goldman, etc.) follow the same snapshot-or-hold rule but classify under Ledger 3 (ecosystem adoption — enterprise operational adoption evidence), not Ledger 1 / Tier 0(b). Only foundation-model companies get the Tier 0(b) lift; the rationale is that FM-co operational tooling decisions are themselves treated as authoritative signal in the way GAO / NIST PDFs are. A non-FM enterprise JD naming a tool is operational adoption evidence comparable to a code import or vendor whitepaper, which is Ledger 3 territory; it is not third-party media (Ledger 2) and it is not a government / FM-co citation (Ledger 1).
references/domain-registry.md from this round's confirmed Phase B hits, grouped by class. Empty list is fine and should still be reported, so the harvest step stays visible across rounds.When running a targeted search (not full audit), append new findings to the existing file. Do not overwrite previous results. Mark the date of each search pass.
news-coverage-audit.md is an internal evidence ledger. It is not what anyone outside the project reads. Every round must therefore end with an explicit decision about which confirmed findings belong on the public site, because a Tier 0 discovery that never reaches opensource.html produces no external value.
Run this step after the audit file is written, never before, so that only verified rows propagate.
Select. From this round's newly confirmed rows, take the ones that change the public impact story: Tier 0 government, international-body, standards, or foundation-model-company citations; Tier 1 items from national labs and major research institutions; and any third-party use of a lab artifact as a baseline or as adopted tooling. Skip Tier 3 through Tier 5 unless a card's headline count is now stale.
Map to the impact cards. opensource.html carries one impact-card per project family (PyOD, Agent Auditability Line, ADBench, TrustLLM, Recent Institutional Visibility, TDC, DoxBench). Add each finding as a row in the matching card, using the existing Type | Evidence | Source column shape and the established impact-type badge classes. Create a new card only when a project family has enough confirmed evidence to stand on its own.
Refresh the derived numbers on every surface, then prove it. This step has been skipped or half-done in more than one round, so it is now a table and a gate rather than a reminder. The patent count once read 12 in the audit and 15 on the site at the same time. The bio's geography list stayed at "Europe, Asia, and the Middle East" for a full round after Brazil was already counted in Ledger 1.
news-coverage-audit.md is the source of truth. Every surface below quotes it, so a finding that changes a figure changes all of them in the same commit.
| Surface | What it carries | Check before closing the round |
|---|---|---|
opensource.html | Impact cards per project family, the External-adoption blurb, the patent row, book and course counts | Does a new Tier 0 or Tier 1 row belong in a card? Do the quoted counts still match the audit? |
files/bio.txt | The adoption sentence: institution classes, geography, patent count, foundation-model-company naming | Does the geography list cover every country now in Ledger 1? Is a newly named institution class present? |
index.html | The PyOD one-line summary, plus a PRERENDER:bio block generated from files/bio.txt | Edit files/bio.txt, then run python scripts/prerender_pages.py to regenerate the block. Never hand-edit inside the markers. |
llms.txt |
The reverse direction also holds: when a public claim on the site cannot be traced to a row in the audit, treat that as a defect in the site, not as evidence the audit is incomplete.
| Mode | When | Dimensions to run |
|---|---|---|
| Full audit | Once per semester, before portfolio updates | All 9 news-coverage dimensions (D1-D8 + D10) plus a freshness-gated hook into the standalone [[citation-audit]] skill (see "Cross-skill: citation-audit integration" below) |
| Targeted | After a specific paper acceptance or tool release | Dims 1, 3, 4, 5 scoped to that item, plus D8 when policy or PDF evidence is plausible |
| Quick check | Before grant submissions | Dims 1, 6, 8, 10 (citations, impact, government PDFs, external deep research) |
| Topic monitor | When a trending topic connects to your work | Dim 4 only, focused on that topic |
| Ecosystem check | Before broader-impact statements | Dim 7 (education, code ecosystem, global) |
| PDF deep search | Before tenure materials or when a specific gov report is suspected | Dim 8 only, with candidate PDF list |
| Affiliation audit | Before tenure / promotion, after major citation milestones | Invoke the standalone /citation-audit skill (was D9 here before; split out as its own skill at skills/citation-audit/SKILL.md) |
| External deep research | After automated audit, as a complement pass | Dim 10 (external LLM deep research) |
Every mode above also runs the Standing Lane (Auditable AI and agent auditability), on top of the dimensions listed for that mode. The Dimensions column says what varies by mode; the Standing Lane does not vary. See "Standing Lane: Auditable AI and Agent Auditability" above for the artifact table, the concept vocabulary, and the watch items.
Purpose: Use external deep research tools (ChatGPT Deep Research, Gemini Deep Research, Claude on claude.ai, or similar) as a complement to the automated Dimensions 1-8. These tools have browsing capabilities, PDF reading, and search strategies that differ from Claude Code's WebSearch, and consistently find items the automated audit misses.
Why this matters: In practice, external deep research tools found the NIST AI 100-2e2025 citation, a third FLI AI Safety Index edition, 5 additional patents, and non-English coverage in Korean/German/Spanish that the automated audit missed entirely. These tools are not a substitute for the structured audit (they lack the systematic coverage and verification discipline), but they are a strong complement.
references/search-queries.md for the base query bank, but the prompt should be open-ended ("search broadly and creatively — I do not know where the coverage is").external-research/ in the project root. Name each file by source and date: {source}-{YYYY-MM}.md (e.g., chatgpt-deep-research-2026-04.md, gemini-2026-04.md, claude-2026-04.md). Date the files so future runs know what was already searched and when. Running once per quarter is sufficient; monthly if a major release or conference just happened.news-coverage-audit.md).Bibliometric citation-affiliation audit (formerly Dimension 9 inside this skill) now lives in its own skill at skills/citation-audit/SKILL.md (slash command: /citation-audit). The two skills divide the external-impact landscape:
news-coverage-audit.md.citation-affiliation-audit.md.The "Full audit" mode of this skill hooks citation-audit results into the editorial report so the final news-coverage-audit.md captures both sides of impact evidence. The hook is freshness-gated, not always-on, because a full citation audit takes 30-80 minutes and may not be appropriate every time news-search runs.
When running the "Full audit" Run Mode, perform the following before writing news-coverage-audit.md:
citation-affiliation-audit.md exists at the project root./citation-audit --source both (or --source openalex if no Dimensions credentials) before the news-search full audit." Do not auto-invoke the long citation audit without confirmation.news-coverage-audit.md under a ## Citation Affiliation Evidence (integrated from citation-audit) section. Specifically:
Category | Institution | Country | Your Work Cited | Citing Paper | Year | Source columns.citation-affiliation-audit.md for the canonical separate copy.(stale, regenerate via /citation-audit), and surface the staleness to the user. Do not silently truncate.The hook never re-runs the citation audit silently; that decision belongs to the user. The hook only reads the existing file and integrates its full content into the unified report.
news-coverage-audit.md is the single artifact a tenure / promotion reader or grant reviewer scans for external-impact evidence. Forcing them to follow a link into a separate citation-affiliation-audit.md (which they may not realize exists) creates a gap in the impact story. Truncating to "top 10 hits" loses the long-tail Tier 1 evidence that often matters most for broader-impact narratives (e.g., a niche citation from Capital One or Mayo Clinic adds new domain breadth that a top-10 list might drop). Embedding the full tables verbatim keeps the integrated report self-contained while the standalone citation-affiliation-audit.md remains canonical for incremental re-runs.
The hook is news-search reading citation-audit's output, not citation-audit calling into news-search. This keeps the two skills independent: the citation-audit pipeline (OpenAlex / Dimensions DSL, Tier 0 / Tier 1 regex patterns, per-source coverage merging) does not need to know anything about editorial-coverage discovery. The single integration surface is the citation-affiliation-audit.md file, which both skills agree on as the contract.
editorial_translationpaywall_or_blocked is held for manual verification, not classified from snippet alone.Tier assignment per the tier structure in the Output section. Assign coverage tiers (Tier 0 through Tier 5) only to candidates that pass direct mention; topic-only candidates keep tier: topic-validation from step 1.
Registry harvest status. For each kept coverage row, set registry_status to existing or new after checking the page's registered domain against references/domain-registry.md. Leave registry_status empty on dropped and topic-validation rows.
| The Institutional Adoption section, which is what a model reads when asked about the lab |
| Does it name the new Tier 0? This surface is easy to forget because no human reads it. |
cv/cv-full.tex, cv/cv-1page.tex | Award source links and grant rows | Do the linked pages still name the PI? Dead or non-naming award links are defects, see below. |
citation-affiliation-audit.md | The bibliometric half | Freshness-gated hook, see the citation-audit integration section. |
The gate. scripts/ci_check_site.py carries check_impact_claims_agree, which reads each figure from news-coverage-audit.md and fails the build when a surface disagrees. When this round introduces a figure that will appear on more than one surface, add it to that function's CLAIMS table in the same commit. A number quoted by only one surface needs no row. A number quoted by two surfaces does, because that is where drift starts.
Check the award links while you are here. A link that returns HTTP 200 is not evidence: fetch each award URL on the public surfaces and confirm the page actually names the PI. The 2026-08-13 round found two dead NortonLifeLock links redirecting to a generic corporate blog, and a Capital One Research Award cited to a USC Viterbi news page that returns 200 and contains no occurrence of "Zhao".
Keep the wording calibrated to the evidence. Carry the audit's caveats onto the site rather than dropping them. Attribution that names only a tool and not the PI stays described that way. A co-authored project stays described as co-authored. Adverse coverage, such as a CVE advisory, is evidence of reach and does not belong in a card that reads as endorsement.
Record what was propagated. Add a short "Propagated to site" note to the round's section in news-coverage-audit.md listing which rows reached which cards, so a later reader can tell the difference between a finding that was never surfaced and one that was deliberately held back.