projekt-workload

Projekt — workload & capacity report (read-only)

One job: a deterministic team workload report. The script fetches the four server-side aggregates and computes every number itself — the model spends no tokens on arithmetic and writes nothing.

SK_CORE="${CLAUDE_SKILL_DIR}/../projekt/scripts" · SK="${CLAUDE_SKILL_DIR}/scripts".

Prerequisite — connect once

Run the projekt skill's setup first so the token is resolved and member names are cached (.projekt-run/context.json); this script reads that cache for name resolution and never re-queries the roster:

bash "$SK_CORE/auth_check.sh"      # resolves user + org, writes context
bash "$SK_CORE/context_sync.sh"    # caches projects + members

Run it

# Current ISO week (Mon→Sun), Markdown table to stdout:
python3 "$SK/scripts/workload_report.py"

# Explicit window:
python3 "$SK/scripts/workload_report.py" --from 2026-06-01 --to 2026-06-07

# CSV for a spreadsheet:
python3 "$SK/scripts/workload_report.py" --csv > workload.csv

# Tune the bands (otherwise over = org /capacity/threshold or 100%, under = 50%):
python3 "$SK/scripts/workload_report.py" --over 90 --under 40

# Machine-readable summary:
python3 "$SK/scripts/workload_report.py" --json

It reads GET /workload?date_from=&date_to=, GET /workload/capacity, GET /capacity and GET /capacity/threshold (see references/endpoints.md → Workload & capacity), merges them per user_id, and prints one row per member: assigned · in-progress · done · hours-logged · est. hours · capacity · utilization % · flag. Members above the over-threshold are flagged ⚠️ OVER, below the under-threshold ↓ under, and those with no capacity target — n/a.

Dry-run / apply

There is nothing to apply — this skill is read-only by construction. It makes only GET calls, takes no --apply/--admit flag, and is always safe to re-run. (The dry-run-by-default rule is satisfied: it never mutates.)

Output columns

Column	Source
Assigned / In progress / Done	counts from /workload (open + WIP + completed in the window)
Hours logged	logged hours from /workload over [from,to]
Est. hours	estimated open load from /capacity (falls back to /workload)
Capacity	per-member target hours from /capacity or /workload/capacity
Utilization %	the server's value if given, else 100 × est-or-logged ÷ capacity
Flag	OVER > over-threshold · under < under-threshold · n/a if no capacity

Gotchas (this domain)

• Threshold source. --over wins; else the org's /capacity/threshold; else 100%. The header line states which was used (flag / org / default).
• Field-name drift. The aggregates use varied keys across orgs/versions; the script normalizes common aliases (in_progress/inProgress/wip, hours_logged/logged_hours, …) and tolerates array or {data:[…]}/{members:[…]} envelopes — the same defensive shape context_sync.sh uses.
• No capacity set. Members with a 0/missing capacity target show utilization — and flag n/a rather than dividing by zero or being silently dropped.
• Names. Resolved from .projekt-run/context.json. If you skipped context_sync.sh, rows fall back to user-ids and a warning prints on stderr.
• Window default. Current ISO week, Monday→Sunday, computed locally. Dates must be YYYY-MM-DD; --from after --to exits non-zero.
• 403 cross-org / 429 / 5xx. A PAT is bound to one org (403 → switch org/token, not a bug); 429 and 5xx are auto-retried with backoff by the shared client. See references/errors.md.

What it does NOT do

No writes, no rebalancing, no assignment changes, no estimation, no time logging — it only reports. To act on the findings (assign/move issues, log time, document) route back through the projekt orchestrator to projekt-issues, projekt-time or projekt-docs. Token resolution + error envelope live in the shared references/auth-setup.md and references/errors.md.