fix(workflows): keep the init step's documented ignore_agent_tools default on an explicit null

[jawwad-ali]

Problem

InitStep documents this default in two places:

class docstring — "Because workflows run unattended, the step defaults to --ignore-agent-tools (skip checks for an installed agent CLI)"

field docs — "ignore_agent_tools — Skip checks for the coding agent CLI (defaults to true)."

It implements it as:

ignore_agent_tools = self._resolve_bool(
    config.get("ignore_agent_tools", True), context
)

config.get(key, default) substitutes the default only when the key is absent. A bare ignore_agent_tools: in YAML parses to None, and _resolve_bool(None) → bool(None) → False.

Reproduction on current main (81bf741)

key ABSENT                   -> ignore_agent_tools=True    flag emitted: YES
bare  ignore_agent_tools:    -> ignore_agent_tools=False   flag emitted: NO   <-- bug
explicit true                -> ignore_agent_tools=True    flag emitted: YES
explicit false               -> ignore_agent_tools=False   flag emitted: NO

So --ignore-agent-tools is dropped, specify init re-runs the agent-CLI presence check, and an unattended workflow run fails with "Agent Detection Error" for any integration whose CLI is not installed on the runner — the precise failure the documented default exists to avoid.

A bare key is a normal YAML authoring shape, and it reads as "leave this at its default", which is the opposite of what happened.

Fix

Normalise an explicit null to the documented default before resolving:

raw_ignore_agent_tools = config.get("ignore_agent_tools")
if raw_ignore_agent_tools is None:
    raw_ignore_agent_tools = True
ignore_agent_tools = self._resolve_bool(raw_ignore_agent_tools, context)

This mirrors how the while/do-while steps handle max_iterations — config.get("max_iterations") followed by an explicit if ... is None default, rather than relying on .get's default.

No breaking change. An absent key, an explicit true, and an explicit false all behave exactly as before — a second new test pins that false still opts in to the agent check. The only input whose behaviour changes is the explicit null, which moves from silently inverting the documented default to honouring it.

Verification

• Fail-before / pass-after: test_explicit_null_ignore_agent_tools_keeps_documented_default fails on unpatched src and passes with the fix. tests/test_workflows.py: 21 failed → 20 failed, 838 → 839 passed.
• The remaining 20 are identical to the clean-main baseline I captured on 81bf741 (all Windows symlink-privilege).
• uvx ruff@0.15.0 check src tests → clean

Tests go into the existing TestInitStep, beside test_builds_here_argv_and_bootstraps, which already asserts --ignore-agent-tools is in the argv.

---

Written with assistance from Claude Code. Bug found, reproduced, and verified by me on current main.