PPT审查
- 作者仓库星标 435
- 作者更新于 实时读取
- 作者仓库 nvim
- 领域
- AI 智能
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @benbrastmckie · 未声明 license
- Token 消耗评级
- 较高消耗
- 接入复杂程度
- 需简单配置
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- Shell 执行
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: skill-slide-critic
description: Interactive critique loop for slide presentations. Delegates to slide-critic-agent, presents fin…
category: AI 智能
runtime: 无特殊运行时
---
# skill-slide-critic 输出预览
## PART A: 任务判断
- 适用问题:提示词、Agent 工作流、模型评估或自动化推理。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Context References / Trigger Conditions / Input Parameters”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于提示词、Agent 工作流、模型评估或自动化推理,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Context References / Trigger Conditions / Input Parameters”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、执行终端命令、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文出现了 `/plan`、`/critique`、`/research` 这类斜杠命令;如果你的 Agent 支持命令触发,优先用命令开场,再补充目标和边界。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件、执行终端命令。
先用一个小任务确认它会围绕“Context References / Trigger Conditions / Input Parameters”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: skill-slide-critic
description: Interactive critique loop for slide presentations. Delegates to slide-critic-agent, presents fin…
category: AI 智能
source: benbrastmckie/nvim
---
# skill-slide-critic
## 什么时候使用
- 把 AI / Agent方向的常用动作沉淀成 Agent 可调用的技能 适合处理AI Agent、提示词、模型评估与自动化推理,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查…
- 面向提示词、Agent 工作流、模型评估或自动化推理,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Context References / Trigger Conditions / Input Parameters」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "skill-slide-critic" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Context References / Trigger Conditions / Input Parameters
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件、执行终端命令 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Slide Critic Skill
Interactive critique feedback loop for academic presentations. Delegates to slide-critic-agent for
initial material review, parses the structured critique report, presents findings to the user grouped
by severity tier, collects accept/reject/modify decisions, loops until all issues are addressed, and
produces a final filtered critique report consumable by /plan.
IMPORTANT: This skill implements the skill-internal postflight pattern. After the subagent returns and the interactive loop completes, this skill handles all postflight operations (status update, artifact linking, git commit) before returning.
Context References
Reference (do not load eagerly):
- Path:
.claude/context/formats/return-metadata-file.md- Metadata file schema - Path:
.claude/context/patterns/postflight-control.md- Marker file protocol - Path:
.claude/context/patterns/file-metadata-exchange.md- File I/O helpers - Path:
.claude/context/patterns/jq-escaping-workarounds.md- jq escaping patterns (Issue #1132)
Note: This skill runs delegation then interactive Q&A. Context is loaded by the delegated agent.
Trigger Conditions
This skill activates when:
/critiquecommand with task number input/researchon a task withtask_type: "present:slides"andworkflow_type: "slides_critique"- Present extension is available
- Task has existing slide materials to review (research reports, plans, or assembled slides)
Input Parameters
Required Parameters
task_number- Task number (must exist in state.json with task_type containing "slides")session_id- Session ID from orchestrator
Optional Parameters
focus_categories- Subset of 6 rubric categories to prioritize (e.g., ["Narrative Flow", "Timing Balance"])audience_context- Audience description for calibrating reviewmaterials_to_review- Override default material discovery (array of file paths)
Execution Flow
Stage 1: Input Validation
Validate required inputs:
task_number- Must be provided and exist in state.json- Verify task_type contains "slides" (supports "present:slides", "slides")
# Lookup task
task_data=$(jq -r --argjson num "$task_number" \
'.active_projects[] | select(.project_number == $num)' \
specs/state.json)
# Validate exists
if [ -z "$task_data" ]; then
return error "Task $task_number not found"
fi
# Extract fields
task_type=$(echo "$task_data" | jq -r '.task_type // ""')
status=$(echo "$task_data" | jq -r '.status')
project_name=$(echo "$task_data" | jq -r '.project_name')
description=$(echo "$task_data" | jq -r '.description // ""')
forcing_data=$(echo "$task_data" | jq -r '.forcing_data // {}')
# Validate task_type (supports "present:slides" or legacy "slides")
if [ "$task_type" != "present:slides" ] && [ "$task_type" != "slides" ]; then
return error "Task $task_number is not a slides task (task_type=$task_type)"
fi
# Extract talk_type from forcing_data
talk_type=$(echo "$forcing_data" | jq -r '.talk_type // "CONFERENCE"')
Stage 2: Preflight Status Update
Update task status to researching BEFORE invoking subagent.
padded_num=$(printf "%03d" "$task_number")
task_dir="specs/${padded_num}_${project_name}"
mkdir -p "$task_dir"
# Update state.json
jq --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
--arg sid "$session_id" \
'(.active_projects[] | select(.project_number == '$task_number')) |= . + {
status: "researching",
last_updated: $ts,
session_id: $sid
}' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
# Update TODO.md marker to [RESEARCHING]
Create postflight marker:
cat > "${task_dir}/.postflight-pending" << EOF
{
"session_id": "${session_id}",
"skill": "skill-slide-critic",
"task_number": ${task_number},
"operation": "slides_critique",
"reason": "Postflight pending: status update, artifact linking, git commit",
"created": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
"stop_hook_active": false
}
EOF
Stage 3: Prepare Delegation Context
Discover materials to review (if not provided via materials_to_review):
# Auto-discover materials in task directory
report_files=$(ls -1 "${task_dir}/reports/"*.md 2>/dev/null)
plan_files=$(ls -1 "${task_dir}/plans/"*.md 2>/dev/null)
# Check for assembled slides
talk_dir="talks/${task_number}_${project_name}"
slide_files=$(ls -1 "${talk_dir}/slides.md" 2>/dev/null)
Build the materials array from discovered files, or use the materials_to_review parameter if provided.
Delegation context:
{
"session_id": "{session_id}",
"delegation_depth": 1,
"delegation_path": ["orchestrator", "critique", "skill-slide-critic", "slide-critic-agent"],
"timeout": 3600,
"task_context": {
"task_number": N,
"task_name": "{project_name}",
"description": "{description}",
"task_type": "present:slides"
},
"workflow_type": "slides_critique",
"forcing_data": {
"talk_type": "{talk_type}",
"materials_to_review": ["{discovered or provided paths}"],
"focus_categories": "{focus_categories or omitted}",
"audience_context": "{audience_context or omitted}"
},
"metadata_file_path": "specs/{NNN}_{SLUG}/.return-meta.json"
}
Stage 4: Invoke Subagent
CRITICAL: Use the Agent tool to spawn the slide-critic-agent. Do NOT use Skill(...).
Tool: Agent (NOT Skill, NOT Plan)
Parameters:
- subagent_type: "slide-critic-agent"
- prompt: [Include full delegation context from Stage 3]
- description: "Critique presentation materials for task {N}"
The slide-critic-agent will:
- Load the critique rubric and priority matrix
- Read all review materials
- Evaluate against 6 rubric categories (or focus_categories)
- Write a critique report to
specs/{NNN}_{SLUG}/reports/{MM}_slide-critique.md - Write metadata to
specs/{NNN}_{SLUG}/.return-meta.json
DO NOT use Skill(...) - this will FAIL. Always use Task.
Stage 4b: Self-Execution Fallback
CRITICAL: If you performed the work above WITHOUT using the Agent tool (i.e., you read files,
wrote artifacts, or updated metadata directly instead of spawning a subagent), you MUST write a
.return-meta.json file now before proceeding. Use the schema from return-metadata-file.md
with status researched.
If you DID use the Agent tool, skip this stage -- the subagent already wrote the metadata.
Interactive Critique Loop (Stages 5-7)
Stage 5: Parse Critique Report
After the subagent returns, read the critique report and extract findings.
Read metadata file to get the critique report path:
metadata_file="${task_dir}/.return-meta.json" meta_status=$(jq -r '.status' "$metadata_file") critique_report_path=$(jq -r '.artifacts[0].path // ""' "$metadata_file") findings_count=$(jq -r '.metadata.findings_count // {}' "$metadata_file")If agent failed or no findings: Skip the interactive loop. Proceed directly to postflight with the agent's status.
Read the critique report and parse findings using these patterns:
Per-slide heading:
### Slide N ({slide_type})Finding line:- [{severity}] {category}: {description}Suggestion line:Suggested improvement: {text}(indented under finding) General heading:### General (Cross-Cutting)Recommendation tiers:### Must Fix,### Should Fix,### Nice to FixBuild numbered issue list:
issues = [ { id: 1, slide: "Slide 3", severity: "Critical", category: "Narrative Flow", description: "...", suggestion: "...", tier: "Must Fix" }, { id: 2, slide: "Slide 7", severity: "Critical", category: "Audience Alignment", description: "...", suggestion: "...", tier: "Must Fix" }, { id: 3, slide: "General", severity: "Major", category: "Timing Balance", description: "...", suggestion: "...", tier: "Should Fix" }, ... ]Assign tier based on severity:
- Critical -> "Must Fix"
- Major -> "Should Fix"
- Minor -> "Nice to Fix"
If no findings found in report (agent found no issues): Report success with zero findings. Skip interactive loop.
Stage 6: Interactive Critique Loop
Present all findings grouped by severity tier in a single consolidated AskUserQuestion. Collect user decisions and loop until all issues are addressed or user exits.
AskUserQuestion format:
Critique findings for your {talk_type} presentation ({N} total issues):
=== MUST FIX ({count}) ===
1. [Critical] {category} - {slide}: {description}
Suggested: {suggestion}
2. [Critical] {category} - {slide}: {description}
Suggested: {suggestion}
=== SHOULD FIX ({count}) ===
3. [Major] {category} - {slide}: {description}
Suggested: {suggestion}
4. [Major] {category} - {slide}: {description}
Suggested: {suggestion}
=== NICE TO FIX ({count}) ===
5. [Minor] {category} - {slide}: {description}
Suggested: {suggestion}
---
For each issue, respond with its number and action:
1: A (accept as-is)
3: R (reject/dismiss)
5: M add comparison to Smith 2024 (modify the suggestion)
Shortcuts: "accept all", "reject all minor", "done" (accept remaining)
Response parsing grammar:
Parse user response line by line. Each line matches one of:
| Pattern | Action | Effect |
|---|---|---|
{N}: A |
Accept issue N | Mark as accepted, use original suggestion |
{N}: R |
Reject issue N | Mark as rejected/dismissed |
{N}: M {text} |
Modify issue N | Mark as modified, store user's text |
accept all |
Bulk accept | Accept all unaddressed issues |
reject all minor |
Bulk reject minor | Reject all Minor severity issues |
reject all |
Bulk reject | Reject all unaddressed issues |
done |
Finish | Accept all remaining unaddressed issues |
Track decisions per issue:
decisions = {
1: { action: "accepted", modification: null },
3: { action: "rejected", modification: null },
5: { action: "modified", modification: "add comparison to Smith 2024" },
...
}
Loop continuation:
After processing responses:
- If all issues are addressed (each has an accepted/rejected/modified decision): proceed to Stage 7
- If unaddressed issues remain AND user did not say "done": re-present ONLY unaddressed issues in a follow-up AskUserQuestion
- Maximum 3 loop iterations to prevent infinite cycles. After 3 iterations, auto-accept all remaining unaddressed issues.
Stage 7: Generate Filtered Critique Report
Write the final filtered report incorporating user decisions.
Determine artifact number:
next_num=$(jq -r --argjson num "$task_number" \
'.active_projects[] | select(.project_number == $num) | .next_artifact_number // 2' \
specs/state.json)
# Use next available number for the filtered report
filtered_num=$(printf "%02d" "$next_num")
Write to specs/{NNN}_{SLUG}/reports/{MM}_filtered-critique.md:
# Filtered Critique Report: {title}
- **Task**: {N} - {description}
- **Talk Type**: {talk_type}
- **Original Findings**: {total} ({critical} critical, {major} major, {minor} minor)
- **Accepted**: {accepted_count} ({modified_count} with modifications)
- **Rejected**: {rejected_count}
## Accepted Issues
### Must Fix ({count})
- Slide {N}: [{severity}] {category}: {description}
Suggested: {original suggestion}
{If modified: "User modification: {user text}"}
### Should Fix ({count})
- Slide {N}: [{severity}] {category}: {description}
Suggested: {original suggestion}
### Nice to Fix ({count})
- Slide {N}: [{severity}] {category}: {description}
Suggested: {original suggestion}
## Rejected Issues ({count})
{Listed for reference but marked as dismissed}
- Slide {N}: [{severity}] {category}: {description} -- DISMISSED
## User Notes
{Any freeform feedback captured during the loop, or "None"}
## Source Critique Report
Original critique: {critique_report_path}
Update .return-meta.json with final artifact info:
{
"status": "researched",
"artifacts": [
{
"type": "report",
"path": "specs/{NNN}_{SLUG}/reports/{MM}_filtered-critique.md",
"summary": "Filtered critique: {accepted_count} accepted, {rejected_count} rejected of {total} findings"
}
],
"metadata": {
"findings_count": {
"total": N,
"accepted": N,
"rejected": N,
"modified": N
}
}
}
Postflight (ALWAYS EXECUTE)
The following stages MUST execute after work is complete, whether the work was done by a subagent or inline. Do NOT skip these stages for any reason.
Stage 8: Read Metadata File
metadata_file="${task_dir}/.return-meta.json"
if [ -f "$metadata_file" ] && jq empty "$metadata_file" 2>/dev/null; then
meta_status=$(jq -r '.status' "$metadata_file")
artifact_path=$(jq -r '.artifacts[0].path // ""' "$metadata_file")
artifact_type=$(jq -r '.artifacts[0].type // ""' "$metadata_file")
artifact_summary=$(jq -r '.artifacts[0].summary // ""' "$metadata_file")
else
echo "Error: Invalid or missing metadata file"
meta_status="failed"
fi
Stage 9: Update Task Status (Postflight)
| Meta Status | Final state.json | Final TODO.md |
|---|---|---|
| researched | researched | [RESEARCHED] |
| partial | researching | [RESEARCHING] |
| failed | (keep preflight) | (keep preflight marker) |
Stage 10: Link Artifacts
Add artifact to state.json with summary. Use the two-step jq pattern to avoid Issue #1132.
Update TODO.md: Link artifact per @.claude/context/patterns/artifact-linking-todo.md with field_name=**Report**, next_field=**Description**.
Stage 11: Git Commit
git add -A
git commit -m "task ${task_number}: complete slide critique
Session: ${session_id}"
Stage 12: Cleanup
rm -f "${task_dir}/.postflight-pending"
rm -f "${task_dir}/.postflight-loop-guard"
rm -f "${task_dir}/.return-meta.json"
Stage 13: Return Brief Summary
Success:
Slide critique completed for task {N}:
- Evaluated materials against {category_count} rubric categories
- Talk type: {talk_type}
- Original findings: {critical} critical, {major} major, {minor} minor
- User decisions: {accepted} accepted ({modified} modified), {rejected} rejected
- Filtered report: specs/{NNN}_{SLUG}/reports/{MM}_filtered-critique.md
- Status updated to [RESEARCHED]
- Changes committed with session {session_id}
No findings:
Slide critique completed for task {N}:
- No issues found by slide-critic-agent
- Materials appear well-structured for {talk_type} presentation
- Status updated to [RESEARCHED]
- Changes committed with session {session_id}
Partial:
Slide critique partially completed for task {N}:
- Agent produced findings but interactive loop was interrupted
- Run critique again to resume (existing report preserved)
Error Handling
Task not found
Slide critique error for task {N}:
- Task not found in state.json
- No status changes made
Wrong task type
Slide critique error for task {N}:
- Task is not a slides task (task_type={task_type})
- No status changes made
Metadata file missing after agent
Keep status at preflight level (researching) for resume.
No findings from agent
Not an error. Report "no issues found" as success. Skip interactive loop. Write researched status.
User abandons mid-loop
Accept all remaining unaddressed issues. Proceed to filtered report generation. Note in report that some issues were auto-accepted due to early exit.
Git commit failure
Non-blocking. Log failure but continue.
Return Format
This skill returns a brief text summary (NOT JSON). The JSON metadata is written to the file and processed internally.
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核