UniC333 · UniC333 · Jul 13, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -47,6 +47,12 @@ validation tools. Keep compatibility with both Claude Code and Codex users.
   Never assume the current date from training data.
 - Financial data must come from at least two independent sources when the skill
   requires verification.
+- Before spawning research roles, prove live-data access by opening and reading
+  one current primary source. If that probe fails, stop time-sensitive research
+  and request dated user-provided material; never substitute training knowledge.
+- Current price, shares, market cap, per-share metrics, currency, corporate
+  actions, and ADR ratios must follow the point-in-time contract in
+  `skills/financial-data.md`.
 - Use exact arithmetic tools for market cap, valuation, cross-source checks, and
   scenario analysis:
   `python3 tools/financial_rigor.py ...`
@@ -63,6 +69,8 @@ validation tools. Keep compatibility with both Claude Code and Codex users.
 - Before finishing a skill/tool change, run the relevant syntax or generation
   check. For compatibility changes, run:
   `python3 scripts/sync-codex-skills.py`
+- For research-contract or audit-tool changes, also run:
+  `python3 -m unittest discover -s tests -v`
 - To verify generated Codex artifacts are current without rewriting files, run:
   `python3 scripts/sync-codex-skills.py --check`
   and, when slash prompts are relevant:

diff --git a/codex-prompts/bottleneck-hunter.md b/codex-prompts/bottleneck-hunter.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `bottleneck-hunter` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/bottleneck-hunter/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/deep-company-series.md b/codex-prompts/deep-company-series.md
@@ -1,11 +1,11 @@
 ---
-description: "AI Berkshire slash entry for 深度公司系列：8 篇长文拆一家公司."
+description: "AI Berkshire slash entry for 看懂XX公司（深度公司系列）：3-8 篇长文拆一家公司."
 argument-hint: $ARGUMENTS
 ---
 
 Use the installed AI Berkshire Codex skill `deep-company-series` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/deep-company-series/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/dyp-ask.md b/codex-prompts/dyp-ask.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `dyp-ask` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/dyp-ask/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/earnings-review.md b/codex-prompts/earnings-review.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `earnings-review` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/earnings-review/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/earnings-team.md b/codex-prompts/earnings-team.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `earnings-team` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/earnings-team/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/financial-data.md b/codex-prompts/financial-data.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `financial-data` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/financial-data/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/industry-funnel.md b/codex-prompts/industry-funnel.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `industry-funnel` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/industry-funnel/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/industry-research.md b/codex-prompts/industry-research.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `industry-research` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/industry-research/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/investment-checklist.md b/codex-prompts/investment-checklist.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `investment-checklist` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/investment-checklist/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/investment-research.md b/codex-prompts/investment-research.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `investment-research` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/investment-research/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/investment-team.md b/codex-prompts/investment-team.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `investment-team` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/investment-team/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/management-deep-dive.md b/codex-prompts/management-deep-dive.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `management-deep-dive` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/management-deep-dive/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/news-pulse.md b/codex-prompts/news-pulse.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `news-pulse` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/news-pulse/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/portfolio-review.md b/codex-prompts/portfolio-review.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `portfolio-review` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/portfolio-review/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/private-company-research.md b/codex-prompts/private-company-research.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `private-company-research` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/private-company-research/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/quality-screen.md b/codex-prompts/quality-screen.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `quality-screen` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/quality-screen/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/thesis-drift.md b/codex-prompts/thesis-drift.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `thesis-drift` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/thesis-drift/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/thesis-tracker.md b/codex-prompts/thesis-tracker.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `thesis-tracker` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/thesis-tracker/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-prompts/wechat-article.md b/codex-prompts/wechat-article.md
@@ -5,7 +5,7 @@ argument-hint: $ARGUMENTS
 
 Use the installed AI Berkshire Codex skill `wechat-article` for this request.
 
-If the skill is not already loaded, read and follow `~/ai-berkshire/codex-skills/wechat-article/SKILL.md`.
+If the skill is not already loaded, locate it through the active Codex skill registry or the current repository checkout. Do not assume a fixed home-directory path.
 
 User arguments:
 $ARGUMENTS
diff --git a/codex-skills/bottleneck-hunter/SKILL.md b/codex-skills/bottleneck-hunter/SKILL.md
@@ -8,7 +8,8 @@ description: "AI Berkshire skill: 供应链瓶颈猎手：AI驱动的全球产
 This skill is generated from `skills/bottleneck-hunter.md` so Claude Code and Codex users share one canonical workflow.
 
 - Treat `$ARGUMENTS` as the user's request in the current Codex thread.
-- When the source mentions Claude-only surfaces such as Task, Agent, WebSearch, Bash, Read, or Write, use the closest Codex capability available in this session: subagents when available, web search when needed, shell commands for local tools, and normal file edits for workspace files.
+- The canonical workflow is capability-based rather than tied to a specific client's tool names. Use available subagents when helpful; when concurrency is lower, run roles in waves, and when subagents are unavailable, execute the same role contracts sequentially.
+- Verify live-data access by opening and reading a current primary source before starting time-sensitive research. If live access fails, stop and request dated user-provided material instead of substituting training knowledge.
 - Use shared project tools from `tools/` in this repository. Prefer running commands from the repository root with paths like `python3 tools/financial_rigor.py ...`; if the current thread starts outside the repo, locate the actual checkout path first instead of assuming a fixed home-directory path.
 - Before starting research, run the `date` command to confirm today's date; treat it as the baseline for "latest" data and state the data cutoff date in the report header. Never assume the current date from training data.
 - Preserve the research quality rules from `AGENTS.md`: cross-check financial data, use exact arithmetic tools for valuation/math, and clearly label uncertainty and source gaps.
@@ -116,7 +117,7 @@ Layer 4：
 
 ### 2.3 其他趋势的拆解
 
-对每个确认的超级趋势执行类似拆解。使用WebSearch搜索：
+对每个确认的超级趋势执行类似拆解。使用当前客户端可用的联网检索能力搜索：
 - "{趋势} supply chain bottleneck 2026"
 - "{趋势} shortage critical component"
 - "{趋势} capacity constraint"
@@ -174,9 +175,9 @@ B级瓶颈（有压力）：
 ### 4.1 对每个S级和A级瓶颈，找出所有相关上市公司
 
 搜索方式：
-- WebSearch "{瓶颈环节} supplier listed company"
-- WebSearch "{瓶颈环节} manufacturer stock"
-- WebSearch "{瓶颈产品} market share company"
+- 搜索 "{瓶颈环节} supplier listed company"
+- 搜索 "{瓶颈环节} manufacturer stock"
+- 搜索 "{瓶颈产品} market share company"
 
 ### 4.2 初筛标准（快速过滤）