# Helpdesk > Use after a session that went badly to file a manual triage report — the user-run counterpart to the automatic issue logging (named "Logit" / "Helpdesk"). It asks for explicit consent before sending anything and asks which privacy level to apply (full shares code/data snippets, summary sends nothing identifiable, off sends nothing), captures the issues the agents could not solve on the first attempt, and produces a triage submission that follows the SAME triage process as the automatic path. The deterministic submission builder + privacy redaction + validator live in scripts/helpdesk/logit.py; this skill is the contract. Honest boundary — the actual SEND to the triage server is the server-tier (the SEC handshake + the EVAL server), not part of this in-repo plugin; the skill produces the payload + applies privacy locally. - Skill: `paulingram/helpdesk` (Agent Skill) - Install (CLI): `npx skillmds@latest add paulingram/helpdesk` - Raw SKILL.md: https://api.skillmd.com/api/skills/paulingram/helpdesk/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: paulingram (https://skillmd.com/u/paulingram) - Updated: 2026-09-21 - Page: https://skillmd.com/skills/paulingram/helpdesk --- # Logit / Helpdesk (HD-1 … HD-3) When a session goes badly, the user needs a one-command way to report it for triage — the MANUAL counterpart (HD-1) to the automatic issue logging. This skill captures the report, gets consent, applies the chosen privacy level, and produces a triage submission that the same triage process consumes (HD-3). The deterministic pieces — the submission builder, the privacy redaction, and the validator — live in **`scripts/helpdesk/logit.py`** (stdlib-only, unit-tested). This skill is the contract + the user-consent workflow. Do not re-implement the deterministic pieces in prose — call the module. ## Honest boundary — what is and is NOT in-repo The in-repo skill PRODUCES a privacy-applied triage submission and records consent. The actual **send to the triage server** (HD-2/HD-3) requires the triage server itself — the **SEC handshake** + the **EVAL server** — which is the **server-tier**, designed separately and NOT part of this plugin. So this skill: writes the submission to `.architect-team/helpdesk/.json`, and the transmission is best-effort and server-tier (when a configured triage endpoint exists it is handed over; absent one, the local submission is the deliverable and the user is told the send is pending the server-tier). Do NOT claim a report was "sent to triage" when only the local payload was produced. ## Workflow ### Step 1 — Confirm consent (HD-2) Ask the user explicitly, via `AskUserQuestion`: *"This will send a report of this session to the triage server — are you OK with that?"* If they decline, STOP and produce nothing. Consent is a domain gate (the user's input IS the deliverable), not a process gate — it always fires. ### Step 2 — Choose the privacy level (HD-2 / EVAL-15…17) Ask, via `AskUserQuestion`, which privacy level to apply: - **full** — shares snippets of your code and data with the project (be explicit that identifiable content is included; EVAL-15). - **summary** — keeps ONLY a safe allow-list of structured fields (`summary` / `category` / `what_happened` / `agent_could_not_solve`); ALL other content — code/data snippets, nested objects, unknown keys, non-object items — is dropped (default-deny), so nothing truly identifiable is sent (EVAL-16). - **off** — send nothing (no submission is produced; EVAL-17). ### Step 3 — Capture the report Capture the session summary + the **issues the agents could not solve on the first attempt** (SR-3). For each issue, record `{category, what_happened, agent_could_not_solve}`. Gather `evidence` items only when the level is `full` (or as redactable items the engine will strip under `summary`). **Privacy is default-deny under `summary`** — only the safe allow-list of structured fields survives; code/data/nested/unknown keys are dropped, and the validator REJECTS any non-allow-listed key as a backstop. But the retained free-text fields are kept verbatim, so **do NOT paste identifiable content** (a stack trace, an API key, an email, a file path) INTO `summary` / `what_happened` under `summary` mode — put such content in a dedicated evidence key (e.g. `code_snippet`) so the redactor drops it, or choose `full` if the user consented to share it. ### Step 4 — Build the submission ```bash $(command -v python3 || command -v python) scripts/helpdesk/logit.py \ build --input report.json --privacy --version --consent --out .architect-team/helpdesk/.json ``` `build_submission(...)` applies the privacy redaction, gates on consent, stamps the `version` (EVAL-8) and `source: manual-helpdesk` so the triage process treats it identically to the automatic path (HD-3), and returns `None` for `off`. ### Step 5 — Validate + hand off (HD-3) `validate_submission(...)` confirms consent + version + that a `summary` submission leaks no identifiable data. Then hand the submission to the triage process — when a triage endpoint is configured, transmit (server-tier); otherwise the local submission is the deliverable and the send is pending the server-tier. State plainly which happened — produced-locally vs sent. ## Cross-references - `scripts/helpdesk/logit.py` — the deterministic builder + redaction + validator (the machine). - `commands/logit.md` — the `/architect-team:logit` user entry point. - `scripts/notify/notify.py` — the existing best-effort emit pattern the transmission step mirrors (when a triage endpoint is configured). - CT6-6 §9 (EVAL) + §12 (SEC) — the automatic logging + the triage-server handshake this manual path feeds (the server-tier, designed separately).