alphacouncil-agent 0.5.0 → 0.5.1

package/.claude/commands/alphacouncil-market.md

@@ -0,0 +1,23 @@
+---
+description: What story the market is telling itself, checked against the tape
+argument-hint: [optional theme or sector]
+---
+
+# AlphaCouncil — market narrative
+
+Focus, if given: **$ARGUMENTS**
+
+Call `get_market_narrative`, adding `extra_queries` when a sector or theme is named.
+
+Read it by the gap, not by the ranking:
+
+- **A theme leading coverage while its series has not moved** — the story is ahead of the
+  data, or the market has stopped listening. Say which you think it is, and why.
+- **A series that moved sharply while its theme is barely covered** — something is happening
+  that has not been narrated yet. This is the most valuable thing on the page.
+
+Headline counts measure attention, never truth. Every claim about a theme must be followed
+by what its market series actually did.
+
+The lexicon is fixed, so a genuinely new narrative lands in `unclassified_headlines`
+instead of being discovered. Read those titles and name the theme by hand if one is there.

package/.claude/commands/alphacouncil-quick.md

@@ -0,0 +1,19 @@
+---
+description: A fast directional read: four analysts, no master bench, no verification
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — quick read
+
+Ticker or question: **$ARGUMENTS**
+
+Run the `quick` preset from `list_council_options`: four analysts and the debate, with no
+master bench and no verification pass. Do not ask which preset to use — the user chose this
+command, which is the answer.
+
+**Say once, in the output, what was skipped:** no master lenses and no cross-verification,
+so the confidence is lower than a standard run and the report should say so rather than
+reading like a full council result.
+
+Start with `screen_ticker`, `get_quote` and `get_options_chain` — they are three calls and
+they establish more than any amount of searching will.

package/.claude/commands/alphacouncil-screen.md

@@ -0,0 +1,21 @@
+---
+description: Mechanical filings screen only — no language model judgment
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — mechanical screen
+
+Ticker: **$ARGUMENTS**
+
+Call `screen_ticker` with `ticker`. The CIK resolves itself.
+
+Report the table exactly as returned, and state plainly:
+
+- **A rule whose inputs are missing is `skipped`, never a pass.** Reporting "6/7 passed"
+  without naming the seventh misrepresents the screen.
+- **Surviving is not a recommendation.** It means the name is worth research time.
+- No language model judgment is involved in this screen. Every elimination names the
+  metric, the measured value and the threshold.
+
+For a non-US ticker the SEC universe has no entry; say which market it is and use
+`market_coverage` to say what that market supports instead of returning nothing.

package/.claude/commands/alphacouncil.md

@@ -0,0 +1,39 @@
+---
+description: Run the full investment council on a ticker or a question
+argument-hint: [ticker or question]
+---
+
+# AlphaCouncil — full council
+
+The user asked for: **$ARGUMENTS**
+
+## Do this in order
+
+1. **Ask which council to run first.** Call `list_council_options` and present the three
+   presets with their seat counts and relative cost. A council is 7 to 44 seats and that
+   range is the user's time and money.
+   **Skip the question if they already answered** — if `$ARGUMENTS` names a roster, says
+   "everything", or says "quick", act on it. A confirmation nobody needed is an interruption.
+
+2. Follow `skills/alphacouncil-agent/SKILL.md` from Stage 0. Do not improvise a shorter
+   workflow: the gates exist because a report that looks finished and skipped the bench is
+   worse than an obviously partial one.
+
+3. **Call the tools instead of searching for numbers they can supply.** `screen_ticker` for
+   filings, `get_quote` for price, `get_options_chain` for positioning, `get_macro_snapshot`
+   for context, `get_news` for dated items. Search is for explanation and for what is not
+   yet filed.
+
+4. Every selected master must report before the run can be complete. `out_of_scope` is a
+   conclusion, not an abstention.
+
+5. Give price bands with the condition attached to each. "The cycle position is undetermined"
+   changes what the bands depend on; it does not excuse leaving them out.
+
+## Never
+
+- Never fill a number from memory when a tool could have supplied it.
+- Never present a rule with missing inputs as a pass — it is `skipped`.
+- Never let social or narrative evidence enter the conclusion on its own.
+
+This is educational software. The final report says so.

package/.claude-plugin/marketplace.json

@@ -6,14 +6,14 @@
   },
   "metadata": {
     "description": "AlphaCouncil plugins for public-equity research.",
-    "version": "0.5.0"
+    "version": "0.5.1"
   },
   "plugins": [
     {
       "name": "alphacouncil-agent",
       "source": "./",
       "description": "Multi-agent equity research: evidence packets, bull/bear debate, portfolio-manager decision.",
-      "version": "0.5.0",
+      "version": "0.5.1",
       "author": {
         "name": "Zhao73"
       },

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "alphacouncil-agent",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "description": "Multi-agent public-equity research workflow for Claude Code: spawns analyst workers, gathers sourced JSON evidence packets, runs bull/bear debate, and produces a portfolio-manager Buy/Overweight/Hold/Underweight/Sell decision with a full report.",
   "author": {
     "name": "Zhao73"
@@ -24,5 +24,6 @@
         "${CLAUDE_PLUGIN_ROOT}/mcp/server.mjs"
       ]
     }
-  }
+  },
+  "commands": "./commands/"
 }

package/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "alphacouncil-agent",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "description": "Host-visible or background public-equity research workflow with shared evidence packets.",
   "author": {
     "name": "Zhao73"

package/.grok/commands/alphacouncil-market.md

@@ -0,0 +1,23 @@
+---
+description: What story the market is telling itself, checked against the tape
+argument-hint: [optional theme or sector]
+---
+
+# AlphaCouncil — market narrative
+
+Focus, if given: **$ARGUMENTS**
+
+Call `get_market_narrative`, adding `extra_queries` when a sector or theme is named.
+
+Read it by the gap, not by the ranking:
+
+- **A theme leading coverage while its series has not moved** — the story is ahead of the
+  data, or the market has stopped listening. Say which you think it is, and why.
+- **A series that moved sharply while its theme is barely covered** — something is happening
+  that has not been narrated yet. This is the most valuable thing on the page.
+
+Headline counts measure attention, never truth. Every claim about a theme must be followed
+by what its market series actually did.
+
+The lexicon is fixed, so a genuinely new narrative lands in `unclassified_headlines`
+instead of being discovered. Read those titles and name the theme by hand if one is there.

package/.grok/commands/alphacouncil-quick.md

@@ -0,0 +1,19 @@
+---
+description: A fast directional read: four analysts, no master bench, no verification
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — quick read
+
+Ticker or question: **$ARGUMENTS**
+
+Run the `quick` preset from `list_council_options`: four analysts and the debate, with no
+master bench and no verification pass. Do not ask which preset to use — the user chose this
+command, which is the answer.
+
+**Say once, in the output, what was skipped:** no master lenses and no cross-verification,
+so the confidence is lower than a standard run and the report should say so rather than
+reading like a full council result.
+
+Start with `screen_ticker`, `get_quote` and `get_options_chain` — they are three calls and
+they establish more than any amount of searching will.

package/.grok/commands/alphacouncil-screen.md

@@ -0,0 +1,21 @@
+---
+description: Mechanical filings screen only — no language model judgment
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — mechanical screen
+
+Ticker: **$ARGUMENTS**
+
+Call `screen_ticker` with `ticker`. The CIK resolves itself.
+
+Report the table exactly as returned, and state plainly:
+
+- **A rule whose inputs are missing is `skipped`, never a pass.** Reporting "6/7 passed"
+  without naming the seventh misrepresents the screen.
+- **Surviving is not a recommendation.** It means the name is worth research time.
+- No language model judgment is involved in this screen. Every elimination names the
+  metric, the measured value and the threshold.
+
+For a non-US ticker the SEC universe has no entry; say which market it is and use
+`market_coverage` to say what that market supports instead of returning nothing.

package/.grok/commands/alphacouncil.md

@@ -0,0 +1,39 @@
+---
+description: Run the full investment council on a ticker or a question
+argument-hint: [ticker or question]
+---
+
+# AlphaCouncil — full council
+
+The user asked for: **$ARGUMENTS**
+
+## Do this in order
+
+1. **Ask which council to run first.** Call `list_council_options` and present the three
+   presets with their seat counts and relative cost. A council is 7 to 44 seats and that
+   range is the user's time and money.
+   **Skip the question if they already answered** — if `$ARGUMENTS` names a roster, says
+   "everything", or says "quick", act on it. A confirmation nobody needed is an interruption.
+
+2. Follow `skills/alphacouncil-agent/SKILL.md` from Stage 0. Do not improvise a shorter
+   workflow: the gates exist because a report that looks finished and skipped the bench is
+   worse than an obviously partial one.
+
+3. **Call the tools instead of searching for numbers they can supply.** `screen_ticker` for
+   filings, `get_quote` for price, `get_options_chain` for positioning, `get_macro_snapshot`
+   for context, `get_news` for dated items. Search is for explanation and for what is not
+   yet filed.
+
+4. Every selected master must report before the run can be complete. `out_of_scope` is a
+   conclusion, not an abstention.
+
+5. Give price bands with the condition attached to each. "The cycle position is undetermined"
+   changes what the bands depend on; it does not excuse leaving them out.
+
+## Never
+
+- Never fill a number from memory when a tool could have supplied it.
+- Never present a rule with missing inputs as a pass — it is `skipped`.
+- Never let social or narrative evidence enter the conclusion on its own.
+
+This is educational software. The final report says so.

package/.opencode/command/alphacouncil-market.md

@@ -0,0 +1,23 @@
+---
+description: What story the market is telling itself, checked against the tape
+argument-hint: [optional theme or sector]
+---
+
+# AlphaCouncil — market narrative
+
+Focus, if given: **$ARGUMENTS**
+
+Call `get_market_narrative`, adding `extra_queries` when a sector or theme is named.
+
+Read it by the gap, not by the ranking:
+
+- **A theme leading coverage while its series has not moved** — the story is ahead of the
+  data, or the market has stopped listening. Say which you think it is, and why.
+- **A series that moved sharply while its theme is barely covered** — something is happening
+  that has not been narrated yet. This is the most valuable thing on the page.
+
+Headline counts measure attention, never truth. Every claim about a theme must be followed
+by what its market series actually did.
+
+The lexicon is fixed, so a genuinely new narrative lands in `unclassified_headlines`
+instead of being discovered. Read those titles and name the theme by hand if one is there.

package/.opencode/command/alphacouncil-quick.md

@@ -0,0 +1,19 @@
+---
+description: A fast directional read: four analysts, no master bench, no verification
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — quick read
+
+Ticker or question: **$ARGUMENTS**
+
+Run the `quick` preset from `list_council_options`: four analysts and the debate, with no
+master bench and no verification pass. Do not ask which preset to use — the user chose this
+command, which is the answer.
+
+**Say once, in the output, what was skipped:** no master lenses and no cross-verification,
+so the confidence is lower than a standard run and the report should say so rather than
+reading like a full council result.
+
+Start with `screen_ticker`, `get_quote` and `get_options_chain` — they are three calls and
+they establish more than any amount of searching will.

package/.opencode/command/alphacouncil-screen.md

@@ -0,0 +1,21 @@
+---
+description: Mechanical filings screen only — no language model judgment
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — mechanical screen
+
+Ticker: **$ARGUMENTS**
+
+Call `screen_ticker` with `ticker`. The CIK resolves itself.
+
+Report the table exactly as returned, and state plainly:
+
+- **A rule whose inputs are missing is `skipped`, never a pass.** Reporting "6/7 passed"
+  without naming the seventh misrepresents the screen.
+- **Surviving is not a recommendation.** It means the name is worth research time.
+- No language model judgment is involved in this screen. Every elimination names the
+  metric, the measured value and the threshold.
+
+For a non-US ticker the SEC universe has no entry; say which market it is and use
+`market_coverage` to say what that market supports instead of returning nothing.

package/.opencode/command/alphacouncil.md

@@ -0,0 +1,39 @@
+---
+description: Run the full investment council on a ticker or a question
+argument-hint: [ticker or question]
+---
+
+# AlphaCouncil — full council
+
+The user asked for: **$ARGUMENTS**
+
+## Do this in order
+
+1. **Ask which council to run first.** Call `list_council_options` and present the three
+   presets with their seat counts and relative cost. A council is 7 to 44 seats and that
+   range is the user's time and money.
+   **Skip the question if they already answered** — if `$ARGUMENTS` names a roster, says
+   "everything", or says "quick", act on it. A confirmation nobody needed is an interruption.
+
+2. Follow `skills/alphacouncil-agent/SKILL.md` from Stage 0. Do not improvise a shorter
+   workflow: the gates exist because a report that looks finished and skipped the bench is
+   worse than an obviously partial one.
+
+3. **Call the tools instead of searching for numbers they can supply.** `screen_ticker` for
+   filings, `get_quote` for price, `get_options_chain` for positioning, `get_macro_snapshot`
+   for context, `get_news` for dated items. Search is for explanation and for what is not
+   yet filed.
+
+4. Every selected master must report before the run can be complete. `out_of_scope` is a
+   conclusion, not an abstention.
+
+5. Give price bands with the condition attached to each. "The cycle position is undetermined"
+   changes what the bands depend on; it does not excuse leaving them out.
+
+## Never
+
+- Never fill a number from memory when a tool could have supplied it.
+- Never present a rule with missing inputs as a pass — it is `skipped`.
+- Never let social or narrative evidence enter the conclusion on its own.
+
+This is educational software. The final report says so.

package/AGENTS.md

@@ -69,6 +69,25 @@ Verified against a real install (grok 0.2.101) rather than from documentation:
   Grok Build the workflow arrives through `AGENTS.md` plus the MCP tools, the same shape
   that works on OpenCode.
 
+## Slash commands
+
+Authored once in `commands/` and generated into each host's directory by
+`scripts/sync-personas.mjs`, so `/alphacouncil` means the same thing everywhere.
+
+| Command | What it does |
+|---|---|
+| `/alphacouncil <ticker or question>` | Full council. Asks which preset first, unless the request already says. |
+| `/alphacouncil-quick <ticker>` | Four analysts, no bench, no verification — and says in the output what was skipped. |
+| `/alphacouncil-screen <ticker>` | Mechanical filings screen only. No language-model judgment. |
+| `/alphacouncil-market [theme]` | What the market is talking about, checked against the tape. |
+
+| Host | Where it reads them |
+|---|---|
+| Claude Code | `commands/` via `.claude-plugin/plugin.json`, plus `.claude/commands/` for a checkout |
+| OpenCode | `.opencode/command/` |
+| Grok Build | `.grok/commands/`, and `.claude/commands/` as a high-priority compatibility source |
+| Codex | `~/.codex/prompts/` is **user-scoped**, so it is installed rather than shipped: `mkdir -p ~/.codex/prompts && cp commands/*.md ~/.codex/prompts/` |
+
 ## Market data coverage
 
 Structured financials come from each market's own regulator, and the pipeline degrades in

package/CHANGELOG.md

@@ -2,6 +2,20 @@
 
 Notable changes per release. Dates are UTC.
 
+## [0.5.1] — 2026-07-26
+
+### Added
+
+- **Slash commands**: `/alphacouncil`, `/alphacouncil-quick`, `/alphacouncil-screen` and
+  `/alphacouncil-market`, authored once and generated into every host's command directory.
+  0.5.0 shipped with no command directory at all, so the plugin could only be reached by
+  @-mention or by describing it in prose.
+
+### Fixed
+
+- The npm package did not carry the command directories, which meant an installed copy and
+  a checkout behaved differently.
+
 ## [0.5.0] — 2026-07-26
 
 Everything below is keyless: no API key, no account, no config file.

package/README.md

@@ -123,6 +123,19 @@ The concise handoff is written to `~/.alphacouncil-agent/runs/<run_id>/user_resp
 The full report is written to `~/.alphacouncil-agent/runs/<run_id>/final_report.md`,
 with analyst Markdown files and `artifact_index.md` in the same run directory.
 
+### Slash commands
+
+| Command | What it does |
+|---|---|
+| `/alphacouncil <ticker or question>` | The full council. Asks which preset first, unless your request already says. |
+| `/alphacouncil-quick <ticker>` | A fast directional read — four analysts, no bench, no verification, and it tells you so. |
+| `/alphacouncil-screen <ticker>` | The mechanical filings screen alone. No language-model judgment in it. |
+| `/alphacouncil-market [theme]` | What the market is talking about, checked against what the tape actually did. |
+
+Available in Claude Code, OpenCode and Grok Build as soon as the plugin is installed. Codex
+keeps its prompts user-scoped, so copy them once:
+`mkdir -p ~/.codex/prompts && cp commands/*.md ~/.codex/prompts/`
+
 ## What It Does
 
 Default stock-analysis runs are full runs, not lite summaries:

package/commands/alphacouncil-market.md

@@ -0,0 +1,23 @@
+---
+description: What story the market is telling itself, checked against the tape
+argument-hint: [optional theme or sector]
+---
+
+# AlphaCouncil — market narrative
+
+Focus, if given: **$ARGUMENTS**
+
+Call `get_market_narrative`, adding `extra_queries` when a sector or theme is named.
+
+Read it by the gap, not by the ranking:
+
+- **A theme leading coverage while its series has not moved** — the story is ahead of the
+  data, or the market has stopped listening. Say which you think it is, and why.
+- **A series that moved sharply while its theme is barely covered** — something is happening
+  that has not been narrated yet. This is the most valuable thing on the page.
+
+Headline counts measure attention, never truth. Every claim about a theme must be followed
+by what its market series actually did.
+
+The lexicon is fixed, so a genuinely new narrative lands in `unclassified_headlines`
+instead of being discovered. Read those titles and name the theme by hand if one is there.

package/commands/alphacouncil-quick.md

@@ -0,0 +1,19 @@
+---
+description: A fast directional read: four analysts, no master bench, no verification
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — quick read
+
+Ticker or question: **$ARGUMENTS**
+
+Run the `quick` preset from `list_council_options`: four analysts and the debate, with no
+master bench and no verification pass. Do not ask which preset to use — the user chose this
+command, which is the answer.
+
+**Say once, in the output, what was skipped:** no master lenses and no cross-verification,
+so the confidence is lower than a standard run and the report should say so rather than
+reading like a full council result.
+
+Start with `screen_ticker`, `get_quote` and `get_options_chain` — they are three calls and
+they establish more than any amount of searching will.

package/commands/alphacouncil-screen.md

@@ -0,0 +1,21 @@
+---
+description: Mechanical filings screen only — no language model judgment
+argument-hint: [ticker]
+---
+
+# AlphaCouncil — mechanical screen
+
+Ticker: **$ARGUMENTS**
+
+Call `screen_ticker` with `ticker`. The CIK resolves itself.
+
+Report the table exactly as returned, and state plainly:
+
+- **A rule whose inputs are missing is `skipped`, never a pass.** Reporting "6/7 passed"
+  without naming the seventh misrepresents the screen.
+- **Surviving is not a recommendation.** It means the name is worth research time.
+- No language model judgment is involved in this screen. Every elimination names the
+  metric, the measured value and the threshold.
+
+For a non-US ticker the SEC universe has no entry; say which market it is and use
+`market_coverage` to say what that market supports instead of returning nothing.

package/commands/alphacouncil.md

@@ -0,0 +1,39 @@
+---
+description: Run the full investment council on a ticker or a question
+argument-hint: [ticker or question]
+---
+
+# AlphaCouncil — full council
+
+The user asked for: **$ARGUMENTS**
+
+## Do this in order
+
+1. **Ask which council to run first.** Call `list_council_options` and present the three
+   presets with their seat counts and relative cost. A council is 7 to 44 seats and that
+   range is the user's time and money.
+   **Skip the question if they already answered** — if `$ARGUMENTS` names a roster, says
+   "everything", or says "quick", act on it. A confirmation nobody needed is an interruption.
+
+2. Follow `skills/alphacouncil-agent/SKILL.md` from Stage 0. Do not improvise a shorter
+   workflow: the gates exist because a report that looks finished and skipped the bench is
+   worse than an obviously partial one.
+
+3. **Call the tools instead of searching for numbers they can supply.** `screen_ticker` for
+   filings, `get_quote` for price, `get_options_chain` for positioning, `get_macro_snapshot`
+   for context, `get_news` for dated items. Search is for explanation and for what is not
+   yet filed.
+
+4. Every selected master must report before the run can be complete. `out_of_scope` is a
+   conclusion, not an abstention.
+
+5. Give price bands with the condition attached to each. "The cycle position is undetermined"
+   changes what the bands depend on; it does not excuse leaving them out.
+
+## Never
+
+- Never fill a number from memory when a tool could have supplied it.
+- Never present a rule with missing inputs as a pass — it is `skipped`.
+- Never let social or narrative evidence enter the conclusion on its own.
+
+This is educational software. The final report says so.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "alphacouncil-agent",
-  "version": "0.5.0",
+  "version": "0.5.1",
   "description": "Multi-agent public-equity research workflow plugin for Codex & Claude Code: sourced evidence packets, bull/bear debate, and a portfolio-manager Buy/Overweight/Hold/Underweight/Sell decision.",
   "type": "module",
   "license": "MIT",
@@ -56,6 +56,10 @@
     "LICENSE",
     "README.md",
     "README.zh-CN.md",
-    "README.ja.md"
+    "README.ja.md",
+    "commands/",
+    ".claude/commands/",
+    ".opencode/command/",
+    ".grok/commands/"
   ]
 }