flow-next-sync

Manual Plan-Sync

Manually trigger plan-sync to update downstream task specs.

Preamble

CRITICAL: flowctl is BUNDLED - NOT installed globally. Define once; subsequent blocks use $FLOWCTL:

FLOWCTL="${CODEX_HOME:-$HOME/.codex}/scripts/flowctl"
[ -x "$FLOWCTL" ] || FLOWCTL="<plugin-root>/scripts/flowctl"   # <plugin-root> = the directory two levels above this skill's SKILL.md file (the harness gave you that file's absolute path when the skill loaded); substitute it literally
[ -x "$FLOWCTL" ] || FLOWCTL=".flow/bin/flowctl"

Input

Arguments: $ARGUMENTS Format: <id> [--dry-run]

• <id> - task ID fn-N-slug.M (or legacy fn-N.M, fn-N-xxx.M) or spec ID fn-N-slug (or legacy fn-N, fn-N-xxx), or a resolvable tracker handle (wor-17 / wor-17.M) that flowctl show maps to the linked spec/task (fn-52.10, R16)
• --dry-run - show changes without writing

Workflow

Step 1: Parse Arguments

REPO_ROOT="$(git rev-parse --show-toplevel 2>/dev/null || pwd)"

Parse $ARGUMENTS for:

• First positional arg = ID
• --dry-run flag = DRY_RUN (true/false)

Validate ID first (handle-recognition rule, R16):

• The id is resolved by flowctl show, not by a prefix check. A session that rejects a resolvable tracker handle as an unknown id — because it gated on "must start with fn-" — has broken this. Route the arg through $FLOWCTL show <ID> --json (Step 3); flowctl's widened resolver (fn-52.10) maps a tracker key (wor-17 / wor-17.M) to its linked spec/task, so a resolvable handle is the existing spec/task, never a new id. /flow-next:sync wor-17 therefore resolves the linked spec.
• If no ID provided: "Usage: /flow-next:sync [--dry-run]"
• If the arg does not resolve via flowctl show (Step 3): "Unknown ID. Use fn-N-slug (spec) / fn-N-slug.M (task), a tracker handle (wor-17), or legacy fn-N, fn-N-xxx."

Detect ID type (use the canonical id from flowctl show):

• Contains . (e.g., fn-1.2, fn-1-add-oauth.2, wor-17.2) -> task ID
• No . (e.g., fn-1, fn-1-add-oauth, wor-17) -> spec ID

Step 2: Validate Environment

test -d .flow || { echo "No .flow/ found. Run flowctl init first."; exit 1; }

If .flow/ missing, output error and stop.

Step 3: Validate ID Exists

$FLOWCTL show <ID> --json

If command fails:

• For task ID: "Task not found. Run flowctl list to see available."
• For spec ID: "Spec not found. Run flowctl specs to see available."

Stop on failure.

Step 4: Find Downstream Tasks

For task ID input:

# Extract spec from task ID (remove .N suffix)
SPEC=$(echo "<task-id>" | sed 's/\.[0-9]*$//')

# Get all tasks in spec
$FLOWCTL tasks --spec "$SPEC" --json

Filter to status: todo or status: blocked. Exclude the source task itself.

For spec ID input:

$FLOWCTL tasks --spec "<spec-id>" --json

1. First, find a source task to anchor drift detection (agent requires COMPLETED_TASK_ID):

   • Prefer most recently updated task with status: done
   • Else: most recently updated task with status: in_progress
   • Else: error "No completed or in-progress tasks to sync from. Complete a task first."

2. Then filter remaining tasks to status: todo or status: blocked (these are downstream).

If no downstream tasks:

No downstream tasks to sync (all done or none exist).

Stop here (success, nothing to do).

Done when

• A source task is anchored — the input task in task mode, or the most recently updated done (else in_progress) task in spec mode — or the run stopped with the documented refusal because neither exists.
• DOWNSTREAM_TASK_IDS holds the spec's todo and blocked tasks with the source task excluded. An empty downstream set stops the run here. A session that spawns the agent with nothing downstream has broken this.

Step 5: Gather glossary + decisions + strategy context

Three extra context types help the agent catch drift the spec text alone can't reveal: project-glossary terms (renames where the old spec used a term whose _Avoid_ alias now appears in code), active decision constraints (current code may touch files mentioned in a decision's Consequences section), and strategic-intent drift (completed task contradicts an active STRATEGY.md track or approach).

GLOSSARY_JSON="$("$FLOWCTL" glossary list --json 2>/dev/null \
  || echo '{"groups":[],"file_count":0,"total_terms":0}')"
DECISIONS_JSON="$("$FLOWCTL" memory list --track knowledge --category decisions --json 2>/dev/null \
  || echo '{"entries":[],"legacy":[],"count":0,"status":"active"}')"
STRATEGY_CONTENT="$("$FLOWCTL" strategy read --json 2>/dev/null || echo '{}')"

All three calls are best-effort — empty defaults keep the agent prompt valid when flowctl returns nothing or fails.

Husk short-circuit — when all three of the following hold, skip the extra context entirely (pass the empty defaults; the agent's husk short-circuit at the top of Phase 3b will skip the whole section):

• GLOSSARY_JSON.total_terms == 0 (glossary missing or husk)
• DECISIONS_JSON.count == 0 (no decision entries)
• STRATEGY_CONTENT.sections_filled == 0 OR STRATEGY_CONTENT == {} (no STRATEGY.md or husk — verify with flowctl strategy status --json | jq '.sections_filled // 0')

When any of the three has signal, pass through all three (untouched) and let the agent run the matching subsection (3b.1 / 3b.2 / 3b.3) and skip the empty ones.

When GLOSSARY_JSON.total_terms == 0 but file_count > 0, every group is a husk. Husks carry no signal for drift detection — pass the JSON through untouched and let the agent skip them.

Done when

• All three variables are populated — with the documented empty default preserved on failure, never dropped. A prompt built with a missing GLOSSARY_JSON / DECISIONS_JSON / STRATEGY_CONTENT key has broken this.
• The husk short-circuit passes the empty defaults through when all three carry no signal, and passes all three untouched when any one has signal.

Step 6: Spawn Plan-Sync Agent

Read the cross-spec flag first — the same single config-leaf read /flow-next:work performs, so a repo that opted into cross-spec propagation (planSync.crossSpec=true) gets the same behavior from a manual /flow-next:sync as from the work-loop auto-trigger. Without this, CROSS_SPEC is unset and plan-sync skips the cross-spec phase entirely — the tool you reach for after big drift silently checks only same-spec tasks:

CROSS_SPEC=$($FLOWCTL config get planSync.crossSpec --json | jq -r '.value')

Build context and spawn via Task tool:

Sync task specs from <source> to downstream tasks.

COMPLETED_TASK_ID: <source task id - the input task, or selected source for spec mode>
FLOWCTL: ${CODEX_HOME:-$HOME/.codex}/scripts/flowctl
SPEC_ID: <spec id>
DOWNSTREAM_TASK_IDS: <comma-separated list from step 4>
DRY_RUN: <true|false>
CROSS_SPEC: <the $CROSS_SPEC value read below — literal "true" or "false", NOT "true|false">

GLOSSARY_JSON: <output of `flowctl glossary list --json` from step 5>
DECISIONS_JSON: <output of `flowctl memory list --track knowledge --category decisions --json` from step 5>
STRATEGY_CONTENT: <output of `flowctl strategy read --json` from step 5>

<if DRY_RUN is true>
DRY RUN MODE: Report what would change but do NOT use Edit tool. Only analyze and report drift.
</if>

Use Task tool with subagent_type: flow-next:plan-sync.

Note: COMPLETED_TASK_ID is always provided - for task-mode it's the input task, for spec-mode it's the source task selected in Step 4.

Done when

• CROSS_SPEC was read from planSync.crossSpec and passed to the agent as the literal string true or false.
• The downstream task-spec edits are made by the spawned flow-next:plan-sync agent. A session that edits downstream task specs directly has broken this.
• Under --dry-run the agent is told to report drift without using Edit, and Step 7 closes with "No files modified."

Step 7: Report Results

After agent returns, format output:

Normal mode:

Plan-sync: <source> -> downstream tasks

Scanned: N tasks (<list>)
<agent summary>

Dry-run mode:

Plan-sync: <source> -> downstream tasks (DRY RUN)

<agent summary>

No files modified.

Error Messages

Rules

• Ignores config - planSync.enabled setting is for auto-trigger only; manual always runs
• Any source status - source task can be todo, in_progress, done, or blocked
• Includes blocked - downstream set includes both todo and blocked tasks
• Reuses agent - spawns existing plan-sync agent, no duplication