AGENT SKILL

industry-research-pack

Guides the production of an industry research pack: a source-priority ladder (official bodies > corporate filings > wire media > research firms; blogs excluded), one-source-one-document packaging with a fixed five-section per-source skeleton, per-bullet source citations, figure-by-figure cross-validation against a named second source, warning marks for single-source figures, and a search-quota-exhaustion fallback. Judgments are rendered AGAINST the user's hypotheses, not in support of them.

License 许可
AGPL-3.0
最近更新
2026-11-14
Stars
★ 25
Forks
9

01 前言 Front Matter

name
industry-research-pack
owner
pm
scope
co-consult
status
active
version
1.0.1
metadata
{ "type": "research", "source": "workspace", "triggers": [ "industry research", "research pack", "market scan", "industry landscape" ] }
relates_to
[ { "type": "composes_with", "skill": "research-analysis" }, { "type": "composes_with", "skill": "evidence-ledger" } ]
description
Guides the production of an industry research pack: a source-priority ladder (official bodies > corporate filings > wire media > research firms; blogs excluded), one-source-one-document packaging with a fixed five-section per-source skeleton, per-bullet source citations, figure-by-figure cross-validation against a named second source, warning marks for single-source figures, and a search-quota-exhaustion fallback. Judgments are rendered AGAINST the user's hypotheses, not in support of them.
last_reviewed
1793923200000
prerequisites
[ "research-analysis", "evidence-ledger" ]

03 SKILL.md 正文

Industry Research Pack

Context

Produce a defensible industry research pack: every material claim carries its source, every figure carries a second source or a visible warning mark, and the conclusion tests the user's hypothesis instead of confirming it.

Origin: co-consult engagement t11/t13 (Projects/co-consult PR #86; filed as T-20261007-013). Related workspace skills: research-analysis (the generic analysis process, no output format) and evidence-ledger (the row-per-claim ledger; this pack is its document-shaped overlay).

When to Use

  • The engagement question is an industry-landscape or market-structure hypothesis that must be tested against sources.
  • The deliverable is a packaged, citable research corpus feeding a report or deck (the report-pack-as-citable-substrate contract: consumer deliverables cite per-source documents per section).

Execution Steps

  1. Source-priority ladder. Select and rank sources in this order; a lower rung may never override a higher one without an explicit note:
    1. Official bodies — regulators, ministries, statistics offices, central banks.
    2. Corporate filings — annual reports, audit disclosures, investor materials.
    3. Wire media — Reuters/Bloomberg/Yonhap-class reporting of official events.
    4. Research firms — paid research (cite the report, not the press summary).
    5. Blogs and aggregators — EXCLUDED; at most use them to locate a higher-rung primary source.
  2. One-source-one-document packaging. Each source becomes exactly ONE document in the pack (<date>-<source-name>.md). Never merge two sources into one document — cross-validation happens BETWEEN documents.
  3. Five-section per-source skeleton. Every source document uses: (a) metadata table (publisher, date, URL, author, ladder rung, access date); (b) summary with one bullet per material claim, each ending [Source: <url>]; (c) figure cross-validation against a named second source — corroborated / deviates (both numbers) / single-source; (d) implication lenses (market, regulation, competition, technology); (e) reliability grade A/B/C with a one-line bias note.
  4. Warning-mark protocol. A figure without a second source carries ⚠ SINGLE-SOURCE inline and in the pack roll-up; the mark is removed only by a real second source, never by a rewrite.
  5. Verify-user-hypotheses rule. Render every judgment AGAINST the user's stated hypothesis: attempt to refute it with the pack. Surviving a test is a finding ("not refuted by N sources"); contradiction by a higher-rung source is reported as contradicted.
  6. Search-quota-exhaustion fallback. On budget exhaustion: stop, mark the pack PARTIAL — quota exhausted at <timestamp>, list covered vs uncovered sections, route the remainder to the next session. Never fill gaps from memory or below-ladder sources.

Output Format

A pack directory of one-source-one-document files plus a roll-up:

  • Per source: the five-section skeleton document (§3).
  • Roll-up: findings ledger with per-figure cross-validation verdicts, warning marks, and the hypothesis verdicts ("not refuted" / "contraded by <source>").
  • Coverage statement: covered vs uncovered sections (PARTIAL marker when the quota fallback fired).

Related Skills

  • research-analysis — the generic analysis process this pack instantiates with a fixed output format.
  • evidence-ledger — the row-per-claim ledger discipline; the pack is its document-shaped overlay (same 2+ independent sources requirement).
  • competitive-intelligence, company-intelligence — per-target deep dives that feed the pack's figure cross-validation.

Usage Examples

  • "Build the Q4 industry pack for the engagement question <hypothesis>"
  • "Cross-validate the market-size figure in <source doc>"

04 同仓库更多 Skills

5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25
5throck
5throck/ai-workspace-standards

i18n-layout

Text layout and encoding guidance: character encoding (UTF-8, legacy Korean code pages, BOM hazards), line endings, RTL/bidi handling, and script-specific font selection (Hangul focus). Use when: encoding is corrupted or mojibake appears, an RTL locale must render correctly, fonts must be chosen for a script, or CRLF/BOM issues surface.

★ 25