Publish Hermes Alice workflow skill pack

README.md

@@ -175,6 +175,7 @@ See:
 
 - [docs/integrations/mcp.md](docs/integrations/mcp.md)
 - [docs/integrations/hermes.md](docs/integrations/hermes.md)
+- [docs/integrations/hermes-skill-pack.md](docs/integrations/hermes-skill-pack.md)
 
 Hermes runtime smoke test:
 
@@ -280,6 +281,7 @@ your AI can only be useful if it can remember, resume, and stay aligned with how
 - [Architecture](ARCHITECTURE.md)
 - [MCP](docs/integrations/mcp.md)
 - [Hermes Guide](docs/integrations/hermes.md)
+- [Hermes Skill Pack](docs/integrations/hermes-skill-pack.md)
 - [Integrations](docs/integrations/importers.md)
 - [OpenClaw Guide](docs/integrations/openclaw.md)
 - [Examples](docs/examples/phase9-command-walkthrough.md)

docs/integrations/hermes-skill-pack.md

@@ -0,0 +1,86 @@
+# Hermes Skill Pack for Alice Workflows
+
+This pack provides Hermes-native skills that guide when and how to call Alice MCP tools.
+
+## What This Pack Includes
+
+Pack location in this repository:
+
+- `docs/integrations/hermes-skill-pack/skills/alice-workflows/`
+
+Skills:
+
+- `alice-continuity-recall`
+- `alice-resumption`
+- `alice-open-loop-review`
+- `alice-explain-provenance`
+- `alice-correction-loop`
+
+## Skill to Tool Map
+
+| Skill | Primary Alice MCP tools |
+|---|---|
+| `alice-continuity-recall` | `alice_recall`, `alice_recent_decisions` |
+| `alice-resumption` | `alice_resume`, `alice_context_pack` |
+| `alice-open-loop-review` | `alice_open_loops`, `alice_recent_changes` |
+| `alice-explain-provenance` | `alice_context_pack`, `alice_recall` |
+| `alice-correction-loop` | `alice_memory_review`, `alice_memory_correct`, `alice_recall`/`alice_resume` |
+
+## Install
+
+From the Alice repository root:
+
+```bash
+export HERMES_HOME="${HERMES_HOME:-$HOME/.hermes}"
+mkdir -p "$HERMES_HOME/skills"
+cp -R docs/integrations/hermes-skill-pack/skills/alice-workflows "$HERMES_HOME/skills/"
+```
+
+## Verify Installation
+
+```bash
+HERMES_HOME="${HERMES_HOME:-$HOME/.hermes}" \
+  ./.venv/bin/hermes skills list --source local
+```
+
+You should see the five `alice-*` skills listed.
+
+To confirm three core skills quickly:
+
+```bash
+HERMES_HOME="${HERMES_HOME:-$HOME/.hermes}" \
+  ./.venv/bin/hermes skills list --source local | \
+  rg "alice-continuity-recall|alice-resumption|alice-open-loop-review"
+```
+
+## Skills vs MCP: Responsibility Split
+
+- Skills: decision policy and workflow instructions (when to call tools, how to format output, what evidence to include).
+- MCP tools: runtime execution and deterministic continuity data retrieval/update.
+- Practical rule: use skills to decide behavior; use MCP tools to fetch or mutate continuity state.
+
+## When Hermes Should Prefer Alice Tools
+
+Use Alice tools instead of inference-only answers when:
+
+- the user asks for prior decisions, commitments, or timeline details
+- the user asks to continue interrupted work with concrete next actions
+- the user asks which blockers/open loops remain
+- the user asks for provenance or evidence behind a claim
+- the user asks to correct stale or incorrect continuity records
+
+Concrete examples:
+
+- "What did we decide about rollout gating last week?" -> prefer `alice_recall`
+- "Resume thread `<uuid>` and give next action plus blockers." -> prefer `alice_resume`
+- "What is still blocked for this project?" -> prefer `alice_open_loops`
+- "Why do you think this is true?" -> prefer `alice_context_pack` + `alice_recall`
+- "This memory is outdated, replace it." -> prefer `alice_memory_review` + `alice_memory_correct`
+
+## Suggested Skill Loading
+
+```bash
+./.venv/bin/hermes -s alice-resumption -s alice-open-loop-review
+```
+
+Add `alice-explain-provenance` for evidence-first responses and `alice-correction-loop` for correction-heavy sessions.

docs/integrations/hermes-skill-pack/skills/alice-workflows/DESCRIPTION.md

@@ -0,0 +1,3 @@
+---
+description: Alice workflow skills for Hermes. These skills route continuity work to Alice MCP tools for recall, resumption, open-loop review, provenance explanation, and correction updates.
+---

docs/integrations/hermes-skill-pack/skills/alice-workflows/alice-continuity-recall/SKILL.md

@@ -0,0 +1,56 @@
+---
+name: alice-continuity-recall
+description: Use Alice MCP recall tools for continuity-grounded answers with provenance when users ask what was decided, what changed, or what to remember.
+version: 1.0.0
+author: Alice
+license: MIT
+metadata:
+  hermes:
+    tags: [alice, continuity, recall, mcp]
+    related_skills: [alice-resumption, alice-explain-provenance]
+---
+
+# Alice Continuity Recall
+
+## Goal
+
+Produce recall answers from Alice continuity records instead of free-form memory.
+
+## Trigger Cues
+
+Use this skill when the user asks:
+- what was decided
+- what happened in a thread/project/person scope
+- what should be remembered from prior work
+
+## Required MCP Tools
+
+- `mcp_<alice_server>_alice_recall`
+- Optional: `mcp_<alice_server>_alice_recent_decisions`
+
+`<alice_server>` is usually `alice_core`.
+
+## Workflow
+
+1. Prefer `alice_recall` over inference-only answers.
+2. Use scope filters when available (`thread_id`, `project`, `person`, `since`, `until`).
+3. Keep `limit` bounded (normally `3` to `10`).
+4. Return summary plus provenance-backed evidence IDs.
+
+## Tool Call Templates
+
+```text
+mcp_alice_core_alice_recall({"query":"<topic>","thread_id":"<uuid>","limit":5})
+```
+
+```text
+mcp_alice_core_alice_recent_decisions({"thread_id":"<uuid>","limit":5})
+```
+
+## Output Contract
+
+Always include:
+- direct answer
+- top evidence items (`id`, `title`, `object_type`)
+- provenance notes from the returned item fields
+- uncertainty note if evidence is weak or absent

docs/integrations/hermes-skill-pack/skills/alice-workflows/alice-correction-loop/SKILL.md

@@ -0,0 +1,63 @@
+---
+name: alice-correction-loop
+description: Run deterministic correction workflows with Alice MCP review and correction tools, then verify that future outputs reflect the update.
+version: 1.0.0
+author: Alice
+license: MIT
+metadata:
+  hermes:
+    tags: [alice, continuity, correction, review, mcp]
+    related_skills: [alice-explain-provenance, alice-open-loop-review]
+---
+
+# Alice Correction Loop
+
+## Goal
+
+Apply corrections through Alice review tools and confirm that recall/resumption behavior updates accordingly.
+
+## Trigger Cues
+
+Use this skill when the user asks:
+- this is outdated or wrong
+- correct this memory
+- supersede or mark stale
+
+## Required MCP Tools
+
+- `mcp_<alice_server>_alice_memory_review`
+- `mcp_<alice_server>_alice_memory_correct`
+- Verification: `mcp_<alice_server>_alice_recall` or `mcp_<alice_server>_alice_resume`
+
+`<alice_server>` is usually `alice_core`.
+
+## Workflow
+
+1. Fetch review queue or detail with `alice_memory_review`.
+2. Select correction action:
+   - `confirm`
+   - `edit`
+   - `delete`
+   - `supersede`
+   - `mark_stale`
+3. Apply correction with `alice_memory_correct`.
+4. Re-run `alice_recall` or `alice_resume` to verify behavior changed.
+5. Report both the correction action and the observed post-correction result.
+
+## Tool Call Templates
+
+```text
+mcp_alice_core_alice_memory_review({"status":"correction_ready","limit":10})
+```
+
+```text
+mcp_alice_core_alice_memory_correct({"continuity_object_id":"<uuid>","action":"supersede","replacement_title":"<title>","replacement_body":{}})
+```
+
+## Output Contract
+
+Always include:
+- corrected object ID
+- action applied
+- reason
+- post-correction verification result

docs/integrations/hermes-skill-pack/skills/alice-workflows/alice-explain-provenance/SKILL.md

@@ -0,0 +1,56 @@
+---
+name: alice-explain-provenance
+description: Explain why an answer is trustworthy by referencing Alice continuity provenance and correction state from MCP tool outputs.
+version: 1.0.0
+author: Alice
+license: MIT
+metadata:
+  hermes:
+    tags: [alice, continuity, provenance, explainability, mcp]
+    related_skills: [alice-continuity-recall, alice-correction-loop]
+---
+
+# Alice Explain Provenance
+
+## Goal
+
+Provide evidence-backed explanations for Alice-based answers.
+
+## Trigger Cues
+
+Use this skill when the user asks:
+- why are you saying this
+- what source supports this
+- can you show evidence or provenance
+
+## Required MCP Tools
+
+- `mcp_<alice_server>_alice_context_pack`
+- Optional: `mcp_<alice_server>_alice_recall`
+
+`<alice_server>` is usually `alice_core`.
+
+## Workflow
+
+1. Start from `alice_context_pack` for a scoped evidence set.
+2. If needed, run a focused `alice_recall` query for missing evidence.
+3. Explain answer claims by citing returned continuity object IDs and provenance fields.
+4. If provenance is thin, state uncertainty and propose the next validating step.
+
+## Tool Call Templates
+
+```text
+mcp_alice_core_alice_context_pack({"thread_id":"<uuid>","recent_decisions_limit":5,"recent_changes_limit":5,"open_loops_limit":5})
+```
+
+```text
+mcp_alice_core_alice_recall({"thread_id":"<uuid>","query":"<claim>","limit":5})
+```
+
+## Output Contract
+
+Always include:
+- claim
+- supporting object IDs
+- provenance summary from returned records
+- confidence posture (`high`, `medium`, `low`) based on evidence quality

docs/integrations/hermes-skill-pack/skills/alice-workflows/alice-open-loop-review/SKILL.md

@@ -0,0 +1,55 @@
+---
+name: alice-open-loop-review
+description: Use Alice open-loop MCP tools to review unresolved commitments and prioritize concrete next actions.
+version: 1.0.0
+author: Alice
+license: MIT
+metadata:
+  hermes:
+    tags: [alice, continuity, open-loops, review, mcp]
+    related_skills: [alice-resumption, alice-correction-loop]
+---
+
+# Alice Open-Loop Review
+
+## Goal
+
+Turn open loops into a prioritized action queue grounded in Alice continuity state.
+
+## Trigger Cues
+
+Use this skill when the user asks:
+- what is still open
+- what is blocked or waiting
+- what should I do next from unresolved items
+
+## Required MCP Tools
+
+- `mcp_<alice_server>_alice_open_loops`
+- Optional: `mcp_<alice_server>_alice_recent_changes`
+
+`<alice_server>` is usually `alice_core`.
+
+## Workflow
+
+1. Call `alice_open_loops` for the relevant scope.
+2. Keep output grouped by posture:
+   - `waiting_for`
+   - `blocker`
+   - `stale`
+   - `next_action`
+3. Surface top priorities first (blocked and overdue items before low-risk follow-ups).
+4. Convert each group into explicit next steps.
+
+## Tool Call Template
+
+```text
+mcp_alice_core_alice_open_loops({"thread_id":"<uuid>","limit":10})
+```
+
+## Output Contract
+
+Always include:
+- grouped open-loop summary
+- top 3 priority actions
+- per-item rationale tied to returned loop posture