| name | handoff-append |
| description | The writer contract for the handoff log: how every pipeline agent appends records to .scratch/handoff.jsonl through scripts/handoff.py. Load when producing any handoff record — dispatch-start, a substantive record, or a consultation. The routing side lives in handoff-routing. |
| compatibility | ["claude-code","github-copilot","opencode","junie-cli"] |
| metadata | {"version":"1.1","author":"team"} |
The One Sanctioned Write
All writes to .scratch/handoff.jsonl go through scripts/handoff.py (Python 3 stdlib). Writing the file directly — shell redirection onto it (>>, cat >>, echo >>, tee), or a Write/Edit tool call — is prohibited. It skips validation and corrupts the log: a missing trailing newline glues two records onto one line, and the file stops parsing. Feeding a record to append on stdin is distinct from a raw write: a heredoc piped into append is the sanctioned input mechanism.
Append the record by piping it to stdin through a quoted heredoc — the quotes (<<'EOF') keep the body literal so nothing in the JSON is shell-expanded:
python3 scripts/handoff.py append <type> <<'EOF'
{"type":"<type>","req_id":"<req-id>","author":"<agent>", …}
EOF
The heredoc must sit directly on the python3 command, exactly as above. The command line must begin with python3 scripts/handoff.py append <type>. Do not wrap it as cat <<'EOF' | python3 …, and do not feed the record with echo …|, printf …|, or cat file |. A producer before the pipe makes the command start with cat/echo/printf, which no tool's auto-allow matches — the append then stalls on a permission prompt. The temptation is strongest for large records, where cat <<'EOF' | looks convenient. The direct heredoc above is the one shape every tool pre-approves, at any record size.
Validation and the Receipt
append stamps ts from the system clock — never compose one; a supplied value is overwritten. Ledger time is wall-clock time by construction, which the board's durations and cost windows depend on. It then validates the stamped record against schemas/scratch/<type>.schema.json before writing. An invalid record is rejected with the schema error — fix the record, never the file. Accepted records are written in canonical form: fields in schema declaration order (type, req_id, ts, author first, payload next, optional fields last), one record per line, newline-terminated.
On success append prints the new record's line number — use it for later responding_to and in_response_to references.
Append-Only Discipline
Never edit, reorder, or delete a prior record. If a prior record has a mistake, append a new record that supersedes it (supersedes_record_at where the schema carries it, a fresh record otherwise). Prior records are the audit trail; the latest record per (req_id, type) is the active state.
Dispatch-Start (First Tool Call)
Every dispatched project-defined agent except pipeline-coordinator and the terminal change-grader appends one dispatch-start record as its first tool call, right after its Scoping Pre-Check sentences (where the dispatch runs one). Skipping it leaves the harness blind to the dispatch's outcome: the record is the start half of the dispatch-event contract, and truncation detection keys on it (handoff-routing skill § Dispatch Truncation Detection).
python3 scripts/handoff.py append dispatch-start <<'EOF'
{"type":"dispatch-start","req_id":"<active req>","author":"<your agent name>","responding_to":[<line>]}
EOF
author is your agent name. responding_to lists the 1-indexed line number(s) of the inbound record(s) the dispatch responds to — your agent definition names the typical anchor; use [0] when no inbound record exists (fresh intake).
Writer Commands
| Operation | Command |
|---|
| Append a record | python3 scripts/handoff.py append <type> — record JSON on stdin (canonical form above) |
| Next retry counter | python3 scripts/handoff.py next-retry --req-id <id> — build-failure records for the req_id after the latest design-block line, plus one |
Anchor a response (responding_to, in_response_to) | python3 scripts/handoff.py latest --type <type> [--req-id <id>] |
| Whole-file check | python3 scripts/handoff.py validate |
| Human inspection (raw records) | python3 scripts/handoff.py show [--last N] |
| Slice board | python3 scripts/handoff.py view [--req-id <id>] — the handoff-board skill |
Reading the whole log with the Read tool for context is fine. Routing decisions belong to the router — route and the coordinator (handoff-routing skill); writers read to anchor their own records, not to route.
Exit codes: 0 success, 1 validation or parse error, 2 usage error, 3 no matching record. view exits 0 on a missing or dirty log and 3 only for --req-id with no records. Two engine-authored records bypass the append command under their own determinism contracts: grading.py extract appends grader-features, and grading.py review-plan appends the engine's review-plan. One record is root-appended on the human's behalf: the consultation-response closing a human-consultation halt — author: "human", in_response_to the request's line.
Permission Setup (One-Time, Per Tool)
append is the pipeline's only sanctioned write, and it is safe — append-only, schema-validated, scoped to the log. Each tool's permission layer must pre-approve it so routine appends do not prompt; the agent's tool grant alone does not.
- Claude Code — pre-approved by a committed
PreToolUse hook (.claude/hooks/handoff-allow.py, registered in .claude/settings.json). It auto-allows python3 scripts/handoff.py invocations and defers everything else; a prefix allow-rule cannot cover the heredoc form, so the hook is required. A companion guard (.claude/hooks/handoff-log-guard.py) denies raw writes to the log — Write/Edit tool calls on it and shell redirection onto it.
- OpenCode — pipeline agents already declare
bash: allow, which runs the command without a prompt; no extra setup.
- Copilot CLI — launch with
--allow-tool 'shell(python3:*)', or add a preToolUse hook to its user config.json.
- Junie — add an allowlist rule
{ "pattern": "python3 scripts/handoff.py **", "action": "allow" } to ~/.junie/allowlist.json, or run in brave mode.
Only Claude Code supports a committed deny on raw writes; the other tools shape the path by pre-approving the sanctioned form alone. The cross-tool backstop is deterministic detection: the quality gate runs python3 scripts/handoff.py validate (see the code-quality-gate skill), so a raw write that corrupts the log fails the gate before review on every tool.
