Personal Profile Generator
Purpose
Generate an evidence-backed personal work profile by collecting multi-source work signals from the current user's authorized workspace, extracting concise traits, saving reusable profile memory when possible, and delivering a readable report plus an optional infographic prompt.
It is designed for portable Feishu/Lark CLI workflows across different users, tenants, and workspaces.
Workflow
- Establish the user's environment:
- Locate the available Feishu/Lark CLI command.
- Check configured profiles and the current authorized user.
- Never assume a hardcoded profile, tenant, app, local path, or person.
- Confirm scope only if missing or risky:
- Target person: default to the current authorized Feishu/Lark user for the selected profile.
- Time range: default to the last 30 days.
- Output: default to a markdown report; add infographic prompt unless the user asks for image generation.
- Collect evidence from available sources:
- Feishu documents and Wiki pages edited, created, searched, or referenced by the user.
- Calendar events, meeting titles, meeting notes, and minutes if accessible.
- Feishu IM groups and messages where permitted.
- Local persistent memory files and prior profile records if present.
- Build an evidence ledger before judging:
- Record source, date, short quote or paraphrase, inferred signal, and confidence.
- Separate observed facts from interpretation.
- Prefer repeated signals across sources over one-off anecdotes.
- Derive three core labels:
工作画像: 7 Chinese characters or fewer.工作风格: 7 Chinese characters or fewer.成长方向: 10 Chinese characters or fewer.
- Generate the report:
- Load
references/profile-schema.md. - Follow its report sections, core trait constraints, evidence ledger, quality bar, and save format.
- For Chinese-facing maintenance, use
references/zh-CN-output-spec.md as the Chinese companion spec.
- Persist the result when a persistence mechanism is available:
- Prefer the user's existing profile store or memory convention.
- If no store exists, create/update a local JSON file only when the user asked for a saved artifact.
- Optionally create or prompt an infographic:
- Use a hand-drawn, warm, structured information-card style.
- Include the three labels, capability tags, evidence highlights, representative scenarios, and track advice.
- If the user asks to generate the image directly, load
references/infographic-generation.md, build the final image prompt from the profile, then call the image generation tool.
Evidence Rules
- Do not invent inaccessible data. Say which sources were unavailable or unauthorized.
- Do not request broad permissions silently. Explain missing scopes and ask before initiating a new authorization flow.
- Do not overfit on a single document, meeting, or message.
- Keep sensitive raw chat excerpts minimal; paraphrase unless the exact wording matters.
- Treat profile labels as hypotheses grounded in evidence, not fixed judgments.
- When data is thin, produce a "low-confidence draft" and list the missing evidence needed.
Feishu CLI Collection
When the task involves Feishu/Lark data, read references/feishu-cli.md for portable command patterns, profile discovery, authorization, and scope checks.
Reference Routing
Load reference files on demand for the current task; do not load every reference by default.
references/feishu-cli.md: portable Feishu/Lark CLI setup, profile selection, authorization, and data collection.references/profile-schema.md: final report sections, core trait constraints, evidence ledger, save format, and infographic data contract.references/infographic-generation.md: hand-drawn image, visual card, poster, infographic, or profile map generation requirements.
Chinese reference versions are available for maintainers and Chinese-speaking users:
references/zh-CN-overview.md: Chinese usage overview and workflow.references/zh-CN-feishu-cli.md: Portable Feishu/Lark CLI setup and data collection guide.references/zh-CN-output-spec.md: Chinese output schema, quality bar, and infographic prompt guide.