Skip to content
sdlc

T-TZD2-task-work-spawn-fast-path-no-op

Status: closed/done · Impact: medium · Complexity: small

Auto-generated from a /sdlc:task-work post-mortem. Review and promote to open/ready before picking up.

Goal

Section titled “Goal”

/sdlc:task-work Step 8 unconditionally spawns a sub-agent to convert post-mortem friction bullets into follow-up task files, even when the friction section reads - none observed. For trivially empty or single-line gap lists this is pure overhead — a round trip through sub-agent dispatch, the spawn-from-post-mortem reference doc, and the no-commit return path. In environments where the Agent tool isn’t exposed (some Skill harness configurations), the runner ends up inline-executing the procedure manually anyway. An inline fast-path keeps the common case cheap and gives the runner a fallback contract when sub-agent dispatch isn’t available. See T-Y0SQ-task-work-emit-completion-marker for the originating post-mortem.

Today

Section titled “Today”

Step 8 of plugin/skills/task-work/SKILL.md (subsection “Spawn follow-up tasks from the friction list”) instructs the runner to “Always invoke this sub-step — even if the gaps section reads - none observed, calling the sub-agent makes the ‘no-op’ explicit and gives one consistent code path.” The reference doc at plugin/entities/task/spawn-from-post-mortem.md handles the no-op case at step 1 (“If the only bullet is - none observed, jump straight to step 7 with a count of zero”), but step 7 still requires a structured stdout report. There is no documented behavior for “the Agent tool isn’t available in this environment” — the runner silently inline-executes the procedure or skips, depending on its own judgment.

Proposed

Section titled “Proposed”

Step 8 gains two explicit shapes:

  1. An inline fast-path: when the runner can statically determine the friction list is empty (only - none observed or no bullets), skip the sub-agent dispatch and emit a one-line report (“SPAWNED: 0, LINKED-EXISTING: 0, SKIPPED: 0”) inline.
  2. A documented fallback contract: when the Agent tool isn’t available, the runner inlines the spawn-from-post-mortem procedure itself, following the same input/output contract. The SKILL.md states this explicitly so the runner doesn’t have to guess.

Approach

Section titled “Approach”
  1. Edit plugin/skills/task-work/SKILL.md Step 8 “Spawn follow-up tasks from the friction list” subsection to add the two shapes above as explicit guidance.
  2. Decide whether the inline fast-path is the runner’s responsibility (parse the friction list, branch) or the sub-agent’s (always dispatched, returns early). Bias toward the runner so the fast-path actually saves the dispatch round trip.
  3. Update plugin/skills/task-work/invariants.yaml if the new prose warrants pinning a phrase (e.g. a literal fallback marker).

Files to touch

Section titled “Files to touch”
  • plugin/skills/task-work/SKILL.md — Step 8 spawn subsection.
  • plugin/skills/task-work/invariants.yaml — possibly pin new fast-path or fallback phrases.

Acceptance criteria

Section titled “Acceptance criteria”
  • AC-1: Step 8’s “Spawn follow-up tasks” prose documents an inline fast-path for - none observed friction lists.
  • AC-2: Step 8’s prose documents what to do when the Agent tool is not available in the runner’s environment.
  • AC-3: lint_skill_prose.py continues to pass against the edited SKILL.md.

Out of scope

Section titled “Out of scope”
  • Building a generic Skill-harness “is tool available” probe — this task only documents the fallback at the prose level.

Dependencies

Section titled “Dependencies”
  • none

Discovery context

Section titled “Discovery context”

Spawned by /sdlc:task-work post-mortem of T-Y0SQ-task-work-emit-completion-marker on 2026-05-21.

Post-mortem

Section titled “Post-mortem”

Captured by /sdlc:task-work on 2026-05-22. PR: pending.

Acceptance criteria coverage

Section titled “Acceptance criteria coverage”
  • AC-1: auto — command grep "Fast-path: empty friction list" against the edited SKILL.md confirms the subsection landed; lint_skill_prose.py passes.
  • AC-2: auto — command grep "Fallback: the .Agent. tool isn't available" confirms the subsection landed; lint_skill_prose.py passes.
  • AC-3: auto — plugin/scripts/lint_skill_prose.py plugin/skills/task-work/SKILL.md reports “1 skill(s) checked, 1 with invariants.yaml, no violations.” against the edited SKILL.md.

What worked

Section titled “What worked”
  • The relevance check found every cited path on first probe; no spec drift between authoring and pickup.
  • The invariants pin (SPAWNED-LOCAL: 0) was a natural extension of the existing spawn-from-post-mortem.md pin — adding a second adjacent invariant kept the shape consistent with what was already there.
  • Quality checks ran cleanly except for a pre-existing audit_entities drift on 2026-05-21-pr-check-schema-promote.md that also fails on main and is unrelated to this change.

Friction and automation gaps

Section titled “Friction and automation gaps”
  • Quality-checks runner conflates pre-existing unrelated drift with new failures — run_quality_checks.py returns FAIL whenever audit_entities.py flags any drifted file, even when none of the drifted files were touched by the current branch. The runner could diff the audit output against the merge-base on origin/main and only fail on drift the current branch introduced; that would let task-work skills proceed without per-run manual triage of pre-existing failures. → spawned T-H69K-run-quality-checks-isolates-pre-existing-drift.

← Back to Tasks