postplus-workspace
Establish reusable brand facts from an official website, use the shared PostPlus brand and file library, and maintain optional local project context. Keep saved brand facts and research separate.
How do I install this agent skill?
npx skills add https://github.com/postplusai/postplus-skills --skill postplus-workspaceIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill 'postplus-workspace' manages project folders and templates for advertising campaigns. It follows good security practices, such as preventing access to sensitive system directories and using normalization to prevent character-based attacks. The only identified risk is a potential for indirect prompt injection if malicious data is placed in the project files, which is a common characteristic of data-processing skills.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
PostPlus Workspace
Start with an official website, supplied brand materials, an existing PostPlus brand, saved files, or an explicitly requested project folder. Reuse confirmed facts across Web, CLI and MCP. A local folder is optional and never required for independent production. The agent owns context selection; the current CLI/server own schemas, revisions, permissions, costing, execution and recovery. Specialist skills own domain judgment.
Core Contract
- Read the selected cloud brand/version or relevant local context before acting.
- Separate confirmed facts, approved claims, research, hypotheses and unknowns.
- Never invent product facts, prices, offers, identities, files or approvals.
- Ask when missing/conflicting information changes truth, intent or execution.
- Make reversible low-risk choices within the user's scope without extra gates.
- Discover current capabilities through the catalog and command help.
- Reuse task/standing spending authorization; confirm changes to approved truth, publication, live advertising budgets and other unapproved external actions.
- Preserve original request identity on retries; retain actual results and cost.
Route
| Request | Route |
|---|---|
| Build or improve a Brand Kit from a website, pasted text or documents | Analyze and save a shared Brand Kit |
| Use a brand or saved files | Read the shared personal library |
| Create a project folder | Initialize local workspace |
| Start a campaign or continue a project | Read stage context, then route |
| Research, strategy, testing or production | Read relevant facts and specialist skill |
| Analyze performance | Record observations, conclusions and next hypothesis |
Analyze And Save A Shared Brand Kit
Read brand-analysis.md for both website and material inputs. It owns the analysis method and mapping to Brand Kit content. The current conversational AI performs the synthesis; website extraction and the guide do not start an additional model or spend generation credits. Cloud brand work requires neither a local folder nor a Python setup check.
- Resolve the intended brand. Reuse its ID from the conversation, or search
postplus library brands --jsonwhen it is unknown. For updates, readpostplus library brand <id> --jsonand retain its complete content/revision. - Read the supplied source:
- Official website: run
postplus library brand-website --url <official-https-url> --json. Usepreview.content,preview.website_contextandpreview.analysis_guidetogether. Extraction may save verified website images; it does not save brand facts. Retain actual page/image failures and coverage limits. - Pasted text or documents: read the material with the current host's supported
tools, then run
postplus library brand-guide --jsonfor the sameanalysis_guide.instructionsand currentanalysis_guide.content_schema. This command supplies guidance; it does not open, upload or parse a file. If a supplied file is unreadable, name the exact unread portion and continue with the sources you can read.
- Official website: run
- Apply the analysis method to the evidence. Go beyond copying the homepage: connect product mechanism and use to benefits, audience situations, objections, proof and voice. Preserve meaningful detail without filling unsupported facts.
- For a requested creation/update, save the in-scope result with
postplus library brand-save-current <id> --revision <n> --file <content.json>. The request authorizes ordinary organization and grounded analysis: do not add a separate review prompt or ask users to approve each field. A request only to inspect, compare or preview ends before saving. Ask narrowly for a material identity/evidence conflict or an unrequested change to approved meaning; unrelated useful work continues. - Read the saved result with
postplus library brand <id> --json. Save makes it usable immediately across Web, CLI and MCP. In MCP, calllibrary_brandfor that same ID to display the editable card. Never claim a save from a prepared payload, a timeout or a preview alone.
MCP uses library_preview_brand_website for the URL route and library_brand_guide for material analysis, then
library_save_current_brand and library_brand. These are the same shared library boundary as CLI/Web. The returned
analysis_guide.content_schema is authoritative for structure; never invent fields because a template contains a useful
concept.
For updates, merge only requested changes into the latest complete content; retain unrelated manual writing, custom
chapters, media and stable IDs/order. Initial creation uses one new brand UUID and revision 0. Retry the exact same ID,
revision and content after an uncertain outcome. On a revision conflict, reread and replay only edits whose original
passages have not changed. If both writers changed a passage, retain both and resolve only that passage with the user.
Normal saves need no separate adoption step. Old brand-save / brand-adopt are only for an explicitly requested
separate draft journey.
Use postplus library brand-use <id> --sections facts,voice,visual --json (or library_resolve_brand) to select
context for later work. Retain the returned brand/version identity with the brief. Choose actual file inputs and roles
separately; organizing a Brand Kit does not generate assets or launch ads.
Read The Shared Personal Library
Search files with library assets; reuse the original permanent postplus-media:// reference instead of reuploading.
Signed preview URLs expire. Archive/restore changes organization only. Permanent deletion needs explicit intent;
unfinished deletion stays discoverable in archived assets and resumes that same deletion. Never delete originals as
routine cleanup or rerun a paid operation because catalogue registration failed. Recover completed Web tasks with
postplus library recover-task <original-task-id>; hosted generation recovery continues its original operation ID
through the media command.
Local brand.md / product.md / audience.md can be read as supplied materials for the analysis route above, or
exported from a chosen snapshot. No implicit bidirectional overwrites. Campaign plans, full research reports and
hypotheses remain project context, not adopted facts.
Initialize An Explicitly Requested Local Workspace
- Resolve the target and inspect existing files. Do not scatter files elsewhere.
- Run
postplus doctor --skill postplus-workspace --json. If Python is missing, follow its concrete action; stop that local initialization if none is given. - Resolve this installed skill root and run:
<python-3-command> <skill-directory>/scripts/workspace.py init \
--target <folder> --project "<project name>" \
--output <folder>/.postplus/workspace-init.json
Optional --brand, --product, --market, --platform use only sourced/user values. The entrypoint creates missing
files and rejects unsafe targets. Fill confirmed facts, preserve UNKNOWN, then run the onboarding check.
Continue A Local Project Or Campaign
Read the nearest project.yaml, applicable AGENTS.md and only stage-relevant files from
context-schema.md. Route to the specialist skill and resume recorded work rather than
recreating it. To initialize a requested campaign, use the same entrypoint's new-campaign with --target, --name,
optional --slug and an explicit --output report. Fill objective, market, platform, offer, CTA, KPI, constraints and
approval state; set current_campaign only for the intended campaign.
Stage Checks And Durable Results
Follow stage-gates.md. Run the workspace entrypoint's check --target <folder> --stage <stage> --output <report.json> before high-impact local execution. Stages: onboarding, research, strategy, test,
production, launch, learn. BLOCKED stops the affected stage; local READY never replaces current prices, balance,
provider validation, authorization or live state. For campaign stages, pass --campaign <intended-campaign-slug> and
confirm the returned campaign matches the intended one before using its readiness result. Use
uncertainty-policy.md for material gaps. Save confirmed results, actual cost, chosen
brand/version and next actions in the relevant cloud record or requested workspace. Explain blockers and the next useful
action concisely. Local templates are in assets/project-template/.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/postplusai/postplus-skills/postplus-workspace">View postplus-workspace on skillZs</a>