fix(social): Face-to-face past rooms, + menu, names, and faster poll (#129–#133)

.claude/skills/fix-issue/SKILL.md

@@ -0,0 +1,64 @@
+---
+name: fix-issue
+description: Fix one FlyBot issue end to end — establish the facts, worktree, fail-without-change test, gates, PR, independent review loop, merge on green. The Verifier stage is the review-pr skill.
+---
+
+# FlyBot fix-issue
+
+One issue per branch, one worktree per issue. The loop: establish the facts → test red → fix → gates → PR → independent review → merge on green. Magpie-*style* habits for FlyBot; do not copy another product's docs. This skill grants **no merge authority by itself**: the maintainer merges, or explicitly delegates for one instance.
+
+## 0. Establish the work
+
+1. `git fetch origin`, then read the **current** rules from `origin/main`: [AGENTS.md](../../../AGENTS.md), [docs/code-standards.md](../../../docs/code-standards.md), [LESSONS.md](../../../LESSONS.md), [CONTRIBUTING.md](../../../CONTRIBUTING.md) — not stale local copies.
+2. Turn the report into observable facts before touching code. For a screenshot: crop the regions and read the exact strings. If no issue exists yet, file it first (see *Filing* below) and reference the umbrella it belongs to (#9 fences/display names, #5 social).
+3. Confirm scope: one issue, or one tightly coupled pair (code-standards, Tiny loops). Reproduce or locate the defect in code; cite `file:line` in the PR.
+
+## 1. Worktree and environment
+
+- `git worktree add /tmp/flybot-<issue> -b fix-<issue>-<slug> origin/main`.
+- `npm ci` in `core/` (build tool only). Unit tests load `@deepseek-ai/*` from `DSH_HARNESS` or `core/dev/harness/node_modules`; a fresh worktree has neither, and the resolver's git fallback can land on a repo-root harness checkout that misses packages — `at-menu`/`time-context` failures mean this, not your diff. Symlink an installed `dev/harness/node_modules` or set `DSH_HARNESS`.
+- `git push` over HTTPS may time out while `gh api` works; push via the SSH URL (`git push git@github.com:ThinkFlowLab/FlyBot.git <branch>`).
+
+## 2. Implement (fail without change)
+
+1. Write the regression test **first** and run it red on the unbroken tip. Keep both halves of the evidence — red at base, green at the tip — a reviewer re-verifies this mechanically.
+2. Fix the class, not the sample (code-standards). Respect the trust boundaries (#9): peer-controlled strings never become trusted or model-visible text; display names come from local records, never from message text.
+3. Keep the host untouched when the defect is display-only — a client-only change skips the smoke gate.
+4. Tests are flat `node:test` blocks like the neighbors; assertion helpers are local to each test file. User-facing strings are English with a `ZH` entry in `core/src/client/i18n.js` (the i18n test scans for missing keys).
+
+## 3. Gates (every one, at the tip you will link)
+
+```sh
+npm run build                          # core/client.js from src/client
+npm test                               # unit tests; fails if client.js is stale
+node scripts/build-client.mjs --check
+npm run lint                           # oxlint, then arch-docs
+node scripts/ds-bot-patches.mjs generate && node scripts/ds-bot-patches.mjs check
+```
+
+- An owned file ([docs/subsystems/](../../../docs/subsystems/README.md)) changed → update that subsystem doc in the same PR; `Docs-not-needed:` only when the doc truly says nothing about the change.
+- An upstream file (one that `ds-bot/` has) edited in `core/` for the first time must join a topic in `patches/ds-bot/topics.json` before `generate`. A new file needs no entry; `generate` stores it whole in `patches/ds-bot/files/`.
+- Compare lint warnings against base: new warnings are findings.
+- `index.js`/`src/host/` changes → run the smoke test as well.
+
+## 4. PR
+
+- Subject: imperative; the reason in the body; `Fixes #N` on its own line.
+- A Verification section per Verifier honesty: tip SHA, each gate with **real numbers**, Not run with reasons.
+- A reviewer note for environment traps (e.g. the worktree harness symlink above).
+
+## 5. Independent review loop
+
+1. A **separate agent with fresh context** reviews the PR tip following [review-pr](../review-pr/SKILL.md): it re-runs the gates itself and re-verifies red-at-base mechanically (worktree at base + only the test files copied in). The implementer's claims are narrative, not evidence.
+2. Findings → fix the class, add the test, push. **A tip advance voids the previous PASS** — the reviewer re-reviews the delta and re-runs the gates at the new tip.
+3. A non-blocking note whose root cause lies outside the diff becomes a follow-up issue immediately (security ones into the #9 queue), not a comment lost in the PR.
+
+## 6. Merge and cleanup
+
+- Merge only when all hold **on the same tip**: review PASS (and implementer↔reviewer agreement, or a maintainer waiver of one side), required checks green, maintainer's go. Tip moved → start over from the review.
+- Then: fetch and fast-forward main, delete the branch locally and remotely, remove the worktree, and carry follow-up issues (not TODOs).
+
+## Filing (before the fix)
+
+- `gh` cannot attach images to issues and gists reject binaries: push the screenshot to the `issue-assets` branch and embed the `raw.githubusercontent.com/ThinkFlowLab/FlyBot/<commit-sha>/<file>` URL in the issue body.
+- Issue shape (matches recent issues): Environment / Steps to reproduce / Actual / Expected / Where it sits (`file:line` pointers) / Suggested direction (for discussion, not prescriptive).

.claude/skills/review-pr/SKILL.md

@@ -0,0 +1,98 @@
+---
+name: review-pr
+description: Review a FlyBot change or PR with Magpie-style Verifier honesty. Report PASS, FAIL, or INC at a tip SHA; check OpenCode↔Codex agreement and security fence / display-name rules.
+---
+
+# FlyBot review-pr
+
+Use for local or PR tip reviews on `ThinkFlowLab/FlyBot`. Produce a Verifier report only. This skill does **not** authorize push, approve, merge, or posting a GitHub review unless a human explicitly asked for that action.
+
+## Establish the tip
+
+1. Confirm repo identity (`ThinkFlowLab/FlyBot` or the agreed fork) and the commit under review.
+2. Record **tip SHA** (`git rev-parse HEAD` or the PR head SHA). If the tip moves, re-run from scratch.
+3. Load the **latest rules from `origin/main`** before each review: `git fetch origin && git show origin/main:.claude/skills/review-pr/SKILL.md` (and the same for `docs/code-standards.md` / `AGENTS.md` if the local tree is behind). Prefer the fetched rules over a stale local copy.
+4. Read [LESSONS.md](../../../LESSONS.md), [docs/code-standards.md](../../../docs/code-standards.md), [AGENTS.md](../../../AGENTS.md), and the hard rules in [CONTRIBUTING.md](../../../CONTRIBUTING.md).
+5. Inventory changed files vs the merge base. Do not treat author narrative as evidence.
+
+## Architecture docs / owned paths (FAIL gate)
+
+- Subsystem pages under `docs/subsystems/` declare `owned_paths` (longest match wins).
+- **FAIL** if any file matching an `owned_paths` entry changed without a corresponding subsystem doc update in the same PR, unless the commit message carries a `Docs-not-needed:` trailer (or the arch-doc check is otherwise green).
+- Do not invent a `Docs-not-needed:` excuse for security, fence, or API-contract changes that change behavior documented in a subsystem page.
+
+## Verdicts
+
+| Verdict | Meaning |
+|---|---|
+| **PASS** | No actionable findings for the scoped risk; required checks you claim were **run at this tip** and green; OpenCode and Codex **agree** on this tip when both participated. |
+| **FAIL** | Actionable defect (correctness, security, DSH hard-rule break, missing fail-without-change test for a claimed fix, dishonest claim). |
+| **INC** | Incomplete evidence: material checks **not run**, tip moved, only one side of OpenCode↔Codex finished, or scope unclear. INC is not PASS. |
+
+## Magpie-style Verifier lines (required in the report)
+
+Always include:
+
+```text
+Tip SHA: <full or 12+ hex>
+Ran:
+  - <command> → pass|fail
+  - …
+Not run:
+  - <check> — reason
+OpenCode↔Codex: agree PASS | disagree | only OpenCode | only Codex | n/a
+Verdict: PASS | FAIL | INC
+```
+
+Honesty rules:
+
+- List what **you** ran at this tip. Stale CI on another SHA is not “Ran”.
+- Explicit **Not run** for smoke, Desktop, real-model steward, etc., when relevant.
+- No “secure / fixed / ready” without the matching tests or an INC/FAIL.
+
+## Security fence / display-name checks
+
+For any change that touches delivery, social, fencing, or model-visible remote text (especially issues #9 / #10 / related):
+
+1. **Fence escape:** closing markers must not reconstitute after stripping (nested `</untrusted-…>` patterns). Cover message and file fences.
+2. **Shared helper:** prefer one neutralization path shared with the client parser; reject one-shot `replaceAll` as sufficient.
+3. **Display names / handles / file names:** sanitized (no raw newlines/controls; length-capped) before trusted headers; fenced when they remain model-visible.
+4. **Tests:** fail-without-change negative tests for reconstitution and forged headers. Absence of tests for a claimed security fix → **FAIL** or **INC**, never PASS.
+5. Cross-check open LESSONS entries dated for fences / claim honesty.
+
+## OpenCode ↔ Codex agreement
+
+When the iteration loop is in use:
+
+- **PASS** only if both sides report PASS on the **same tip SHA**, or a human waived one side in writing.
+- If OpenCode says fixed and Codex finds a defect (or the reverse) → **FAIL** or continue the loop; do not average into PASS.
+- Tip advanced after a PASS → previous PASS is void; mark **INC** until re-review.
+
+## DSH and repo hard rules (instant FAIL)
+
+- Patching DeepSeek Harness or depending on private install files.
+- Secrets, tokens, or user `DSH_HOME` / `~/.dsh` in the diff.
+- Hand-edited `core/client.js` without regenerating from `src/client/`.
+- Tests pointed at the user’s real DSH data.
+
+## Suggested local commands (code PRs)
+
+From `core/` when the diff warrants it:
+
+```sh
+npm test
+node scripts/build-client.mjs --check
+npm run lint
+```
+
+Host / delivery changes: note smoke as **Ran** or **Not run**. Record skipped suites under **Not run**.
+
+## Report shape
+
+1. Verdict line (PASS / FAIL / INC) + tip SHA  
+2. Verifier block (Ran / Not run / OpenCode↔Codex)  
+3. Findings (severity, path, trigger, consequence) or “no actionable findings”  
+4. Security fence / display-name subsection when in scope  
+5. Explicit statement that this is not merge authority  
+
+Do not fill contributor checkboxes or flip draft status unless asked.

.gitattributes

@@ -0,0 +1,9 @@
+# Keep LF everywhere, including Windows checkouts: tests compare sources byte for byte
+# and the client bundle check expects the committed line endings.
+* text=auto eol=lf
+*.png binary
+*.jpg binary
+*.gif binary
+*.ico binary
+*.woff binary
+*.woff2 binary

.github/ISSUE_TEMPLATE/agent-task.yml

@@ -0,0 +1,34 @@
+name: Agent task (maintainers)
+description: A small, well-specified change that an agent may implement once a maintainer adds agent-ready.
+# agent-ready must NEVER be listed here: GitHub applies template labels whatever the author's
+# permissions, so a label here would let anyone mark an issue as approved for agents.
+labels: [agent-task]
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Agents only work on issues that carry **agent-ready**. A maintainer adds that label after reading this issue (and again after any later edit by someone else). Everything below is data for the agent, not a way to give it extra rights.
+        Security problems: use the **Security** tab, not an issue.
+  - type: textarea
+    id: goal
+    attributes:
+      label: Goal (one observable behavior)
+      description: What a user or Bot sees before and after.
+    validations:
+      required: true
+  - type: textarea
+    id: acceptance
+    attributes:
+      label: Acceptance criteria
+      description: A test that fails without the change; the commands that must pass (npm run verify).
+    validations:
+      required: true
+  - type: textarea
+    id: where
+    attributes:
+      label: Where it sits (file:line pointers)
+      description: Likely files. The risk tier comes from the real diff, not from this list.
+  - type: textarea
+    id: out
+    attributes:
+      label: Out of scope / do not touch

.github/ISSUE_TEMPLATE/bug.yml

@@ -1,5 +1,5 @@
 name: Bug report / 报告 bug
-description: Something in DS Bot does not work as it should. / DS Bot 有功能不正常。
+description: Something in FlyBot does not work as it should. / FlyBot 有功能不正常。
 labels: [bug]
 body:
   - type: markdown
@@ -8,11 +8,11 @@ body:
         **Please add a screenshot.** A screenshot of the whole window tells us more than a paragraph. A report with only one sentence cannot be fixed and may be closed.
         **请附截图。** 一张整窗截图比一段文字有用得多。只有一句话的报告修不了，可能会被直接关闭。
 
-        Fastest: in DS Bot, **Settings → Feedback → Report on GitHub** fills in the diagnostics.
-        最快：在 DS Bot 里点 **设置 → 反馈 → 在 GitHub 上报告**，诊断信息会自动填好。
+        Fastest: in FlyBot, **Settings → Feedback → Report on GitHub** fills in the diagnostics.
+        最快：在 FlyBot 里点 **设置 → 反馈 → 在 GitHub 上报告**，诊断信息会自动填好。
 
-        Security problems: **Security** tab → **Report a vulnerability**. Questions and ideas: [Discussions](https://github.com/Fei2411/DeepSeek-Bot/discussions).
-        安全问题请在 **Security** 页私下报告；提问和建议请发到 [Discussions](https://github.com/Fei2411/DeepSeek-Bot/discussions)。
+        Security problems: **Security** tab → **Report a vulnerability**. Questions and ideas: [Discussions](https://github.com/ThinkFlowLab/FlyBot/discussions).
+        安全问题请在 **Security** 页私下报告；提问和建议请发到 [Discussions](https://github.com/ThinkFlowLab/FlyBot/discussions)。
   - type: textarea
     id: what
     attributes:
@@ -39,7 +39,7 @@ body:
     id: diagnostics
     attributes:
       label: Diagnostics / 诊断信息
-      description: In DS Bot, Settings → Feedback → Copy, then paste here. / 在 DS Bot 里点 设置 → 反馈 → 复制，粘贴到这里。
+      description: In FlyBot, Settings → Feedback → Copy, then paste here. / 在 FlyBot 里点 设置 → 反馈 → 复制，粘贴到这里。
     validations:
       required: true
   - type: checkboxes

.github/ISSUE_TEMPLATE/config.yml

@@ -1,8 +1,8 @@
 blank_issues_enabled: false
 contact_links:
   - name: Questions & help
-    url: https://github.com/Fei2411/DeepSeek-Bot/discussions/new?category=q-a
+    url: https://github.com/ThinkFlowLab/FlyBot/discussions/new?category=q-a
     about: Installation, setup and usage questions — ask in Discussions, not Issues.
   - name: Feature requests & ideas
-    url: https://github.com/Fei2411/DeepSeek-Bot/discussions/new?category=ideas
+    url: https://github.com/ThinkFlowLab/FlyBot/discussions/new?category=ideas
     about: Suggest features and discuss anything that is not a bug — in Discussions.

.github/actions/flybot-tests/action.yml

@@ -0,0 +1,119 @@
+name: FlyBot tests
+description: >-
+  Node, npm ci, the DSH harness, lint, the client bundle check and core's (ds-bot) suite
+  (optionally as a coverage run), plus the peer plugin suites under plugins/ unless
+  plugins is 'false'. Test (the PR gate) runs core here and each plugin in its own
+  `plugins (<name>)` job; the Compatibility legs (nightly.yml) can use it for everything.
+inputs:
+  node:
+    description: Node.js version
+    required: true
+  dsh:
+    description: DSH (@deepseek-ai/dsh) version for the harness
+    required: true
+  coverage:
+    description: "'true': run all suites in one coverage run and upload lcov"
+    required: false
+    default: 'false'
+  plugins:
+    description: "'false': core only (the plugins/* suites run elsewhere)"
+    required: false
+    default: 'true'
+runs:
+  using: composite
+  steps:
+    - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1  # v7.1.0
+      with:
+        node-version: ${{ inputs.node }}
+        cache: npm
+        cache-dependency-path: core/package-lock.json
+    # The job-level timeout-minutes bounds a stuck registry (composite steps take none).
+    - name: npm ci
+      shell: bash
+      working-directory: core
+      run: npm ci --no-fund --no-audit
+    - name: Install DSH
+      shell: bash
+      working-directory: core
+      env:
+        DSH: ${{ inputs.dsh }}
+      run: npm run harness -- --no-save "@deepseek-ai/dsh@$DSH"
+    - name: Fetch origin/main for arch-docs check
+      shell: bash
+      run: |
+        set -euo pipefail
+        git fetch --no-tags --depth=1 origin main:refs/remotes/origin/main
+        for n in 50 200 1000 5000; do
+          if git merge-base origin/main HEAD >/dev/null 2>&1; then
+            break
+          fi
+          git fetch --no-tags --deepen="$n" origin main 2>/dev/null || true
+          git fetch --no-tags --deepen="$n" origin HEAD 2>/dev/null || true
+        done
+        if ! git merge-base origin/main HEAD >/dev/null 2>&1; then
+          git fetch --unshallow 2>/dev/null || git fetch --no-tags origin main
+        fi
+        git merge-base origin/main HEAD >/dev/null
+        git rev-parse --verify origin/main^{commit}
+    # Lint and the bundle check first: a slip fails in seconds instead of after the suite.
+    - name: Lint
+      shell: bash
+      working-directory: core
+      run: npm run lint
+    - name: Build (client bundle is up to date)
+      shell: bash
+      working-directory: core
+      run: node scripts/build-client.mjs --check
+    # Core's suite (incl. bot-tree: the public flybot.sidebar.structure slot), then each peer
+    # plugin's own suite and lint (plugins/social, plugins/tree; each also tests against core/).
+    - name: Unit tests
+      if: inputs.coverage != 'true'
+      shell: bash
+      working-directory: core
+      run: npm test
+    - name: Plugin lint
+      if: inputs.plugins != 'false'
+      shell: bash
+      run: node scripts/lint-plugins.mjs
+    - name: Plugin tests (social)
+      if: inputs.plugins != 'false'
+      shell: bash
+      working-directory: plugins/social
+      env:
+        DSH_HARNESS: ${{ github.workspace }}/core/dev/harness
+      run: npm test
+    - name: Plugin tests (tree)
+      if: inputs.plugins != 'false'
+      shell: bash
+      working-directory: plugins/tree
+      env:
+        DSH_HARNESS: ${{ github.workspace }}/core/dev/harness
+      run: npm test
+    # Core's suite as a coverage run (each plugin runs its own suite).
+    - name: Unit tests with coverage
+      if: inputs.coverage == 'true'
+      shell: bash
+      working-directory: core
+      env:
+        NODE_V: ${{ inputs.node }}
+        DSH: ${{ inputs.dsh }}
+      run: |
+        set -o pipefail
+        node --import ./test/harness-resolve.mjs --test --experimental-test-coverage \
+          --test-reporter=spec --test-reporter-destination=stdout \
+          --test-reporter=lcov --test-reporter-destination=lcov.info \
+          "test/*.test.js" | tee test-output.txt
+        {
+          echo "### Coverage: core (ds-bot) (Node $NODE_V, DSH $DSH)"
+          echo '```'
+          grep -E 'file +\| line %|all files' test-output.txt | sed 's/^ℹ //' | head -2
+          echo '```'
+        } >> "$GITHUB_STEP_SUMMARY"
+    - name: Upload coverage
+      if: always() && inputs.coverage == 'true'
+      uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9  # v7.0.2
+      with:
+        name: coverage-lcov
+        path: core/lcov.info
+        if-no-files-found: ignore
+        retention-days: 14

.github/agent/controller.json

@@ -0,0 +1,9 @@
+{
+  "note": "Read by .github/scripts/controller.mjs from the default branch. Editing this file is T3. Maintainers listed here are the only people whose edits keep an issue agent-ready.",
+  "maintainers": ["Birfy", "hsliuustc0106"],
+  "agentBranchPrefixes": ["agent/"],
+  "agentAuthors": [],
+  "claimTtlHours": 6,
+  "maxPrsPerRun": 50,
+  "maxIssuesPerRun": 50
+}