manual_tests.run_fire_tests

Step 3/4 in run_all workflow

Run all manual tests: reset, NOT-fire tests, fire tests, and infinite block tests

Runs all manual hook/rule tests using sub-agents. Use when validating that DeepWork rules fire correctly.

Prerequisites (Verify First)

Before proceeding, confirm these steps are complete:

• /manual_tests.run_not_fire_tests

Instructions

Goal: Runs all 6 'should fire' tests serially with resets between each. Use after NOT-fire tests to verify rules fire correctly.

Run Should-Fire Tests

Objective

Run all "should fire" tests in serial sub-agents to verify that rules fire correctly when their trigger conditions are met without safety conditions.

CRITICAL: Sub-Agent Requirement

You MUST spawn sub-agents to make all file edits. DO NOT edit the test files yourself.

Why sub-agents are required:

1. Sub-agents run in isolated contexts where file changes are detected
2. When a sub-agent completes, the Stop hook automatically evaluates rules
3. You (the main agent) observe whether hooks fired - you do NOT manually trigger them
4. If you edit files directly, the hooks won't fire because you're not a completing sub-agent

NEVER manually run

echo '{}' | python -m deepwork.hooks.rules_check

- this defeats the purpose of the test. Hooks must fire AUTOMATICALLY when sub-agents return.

CRITICAL: Serial Execution

These tests MUST run ONE AT A TIME, with resets between each.

Why serial execution is required:

• These tests edit ONLY the trigger file (not the safety)
• If multiple sub-agents run in parallel, sub-agent A's hook will see changes from sub-agent B
• This causes cross-contamination: A gets blocked by rules triggered by B's changes
• Run one test, observe the hook, reset, then run the next

Task

Run all 6 "should fire" tests in serial sub-agents, resetting between each, and verify that blocking hooks fire automatically.

Process

For EACH test below, follow this cycle:

1. Launch a sub-agent using the Task tool with:

   • model: "haiku"

     - Use the fast model to minimize cost and latency

   • max_turns: 5

     - Prevent sub-agents from hanging indefinitely
2. Wait for the sub-agent to complete
3. Observe whether the hook fired automatically - you should see a blocking prompt or command output
4. If no visible blocking occurred, check the queue:

   ls -la .deepwork/tmp/rules/queue/
   cat .deepwork/tmp/rules/queue/*.json 2>/dev/null
   

   • If queue entries exist with status "queued", the hook DID fire but blocking wasn't visible
   • If queue is empty, the hook did NOT fire at all
   • Record the queue status along with the result
5. Record the result - pass if hook fired (visible block OR queue entry), fail if neither
6. Reset (MANDATORY after each test) - follow the reset step instructions:

   git reset HEAD manual_tests/ && git checkout -- manual_tests/ && rm -f manual_tests/test_created_mode/new_config.yml
   deepwork rules clear_queue
   

   See reset.md for detailed explanation of these commands.
7. Check for early termination: If 2 tests have now failed, immediately:
   • Stop running any remaining tests
   • Report the results summary showing which tests passed/failed
   • The job halts here - do NOT proceed with remaining tests
8. Proceed to the next test (only if fewer than 2 failures)

IMPORTANT: Only launch ONE sub-agent at a time. Wait for it to complete and reset before launching the next.

Test Cases (run serially)

Test 1: Trigger/Safety

• Sub-agent prompt: "Edit ONLY

  manual_tests/test_trigger_safety_mode/feature.py

  to add a comment. Do NOT edit the

  _doc.md

  file."
• Sub-agent config:

  model: "haiku"

  ,

  max_turns: 5

• Expected: Hook fires with prompt about updating documentation

Test 2: Set Mode

• Sub-agent prompt: "Edit ONLY

  manual_tests/test_set_mode/module_source.py

  to add a comment. Do NOT edit the

  _test.py

  file."
• Sub-agent config:

  model: "haiku"

  ,

  max_turns: 5

• Expected: Hook fires with prompt about updating tests

Test 3: Pair Mode

• Sub-agent prompt: "Edit ONLY

  manual_tests/test_pair_mode/handler_trigger.py

  to add a comment. Do NOT edit the

  _expected.md

  file."
• Sub-agent config:

  model: "haiku"

  ,

  max_turns: 5

• Expected: Hook fires with prompt about updating expected output

Test 4: Command Action

• Sub-agent prompt: "Edit

  manual_tests/test_command_action/input.txt

  to add some text."
• Sub-agent config:

  model: "haiku"

  ,

  max_turns: 5

• Expected: Command runs automatically, appending to the log file (this rule always runs, no safety condition)

Test 5: Multi Safety

• Sub-agent prompt: "Edit ONLY

  manual_tests/test_multi_safety/core.py

  to add a comment. Do NOT edit any of the safety files (

  _safety_a.md

  ,

  _safety_b.md

  , or

  _safety_c.md

  )."
• Sub-agent config:

  model: "haiku"

  ,

  max_turns: 5

• Expected: Hook fires with prompt about updating safety documentation

Test 6: Created Mode

• Sub-agent prompt: "Create a NEW file

  manual_tests/test_created_mode/new_config.yml

  with some YAML content. This must be a NEW file, not a modification."
• Sub-agent config:

  model: "haiku"

  ,

  max_turns: 5

• Expected: Hook fires with prompt about new configuration files

Results Tracking

Record the result after each test:

Test Case	Should Fire	Visible Block?	Queue Entry?	Result
Trigger/Safety	Edit .py only
Set Mode	Edit _source.py only
Pair Mode	Edit _trigger.py only
Command Action	Edit .txt
Multi Safety	Edit .py only
Created Mode	Create NEW .yml

Queue Entry Status Guide:

• If queue has entry with status "queued" -> Hook fired, rule was shown to agent
• If queue has entry with status "passed" -> Hook fired, rule was satisfied
• If queue is empty -> Hook did NOT fire

Quality Criteria

• Sub-agents spawned: Tests were run using the Task tool to spawn sub-agents - the main agent did NOT edit files directly
• Correct sub-agent config: All sub-agents used

  model: "haiku"

  and

  max_turns: 5

• Serial execution: Sub-agents were launched ONE AT A TIME, not in parallel
• Reset between tests: Reset step was followed after each test
• Hooks fired automatically: The main agent observed the blocking hooks firing automatically when each sub-agent returned - the agent did NOT manually run rules_check
• Early termination on 2 failures: If 2 tests failed, testing halted immediately and results were reported
• Results recorded: Pass/fail status was recorded for each test case
• When all criteria are met, include

  <promise>Quality Criteria Met</promise>

  in your response

Reference

See test_reference.md for the complete test matrix and rule descriptions.

Context

This step runs after the "should NOT fire" tests. These tests verify that rules correctly fire when trigger conditions are met without safety conditions. The serial execution with resets is essential to prevent cross-contamination between tests. Infinite block tests are handled in a separate step.

Job Context

A workflow for running manual tests that validate DeepWork rules/hooks fire correctly.

The run_all workflow tests that rules fire when they should AND do not fire when they shouldn't. Each test is run in a SUB-AGENT (not the main agent) because:

1. Sub-agents run in isolated contexts where file changes can be detected
2. The Stop hook automatically evaluates rules when each sub-agent completes
3. The main agent can observe whether hooks fired without triggering them manually

CRITICAL: All tests MUST run in sub-agents. The main agent MUST NOT make the file edits itself - it spawns sub-agents to make edits, then observes whether the hooks fired automatically when those sub-agents returned.

Sub-agent configuration:

• All sub-agents should use

  model: "haiku"

  to minimize cost and latency
• All sub-agents should use

  max_turns: 5

  to prevent hanging indefinitely

Steps:

1. reset - Ensure clean environment before testing (clears queue, reverts files)
2. run_not_fire_tests - Run all "should NOT fire" tests in PARALLEL sub-agents (6 tests)
3. run_fire_tests - Run all "should fire" tests in SERIAL sub-agents with resets between (6 tests)
4. infinite_block_tests - Run infinite block tests in SERIAL (4 tests - both fire and not-fire)

Reset procedure (see steps/reset.md):

• Reset runs FIRST to ensure a clean environment before any tests
• Each step also calls reset internally when needed (between tests, after completion)
• Reset reverts git changes, removes created files, and clears the rules queue

Test types covered:

• Trigger/Safety mode
• Set mode (bidirectional)
• Pair mode (directional)
• Command action
• Multi safety
• Infinite block (prompt and command) - in dedicated step
• Created mode (new files only)

Required Inputs

Files from Previous Steps - Read these first:

• not_fire_results

  (from

  run_not_fire_tests

  )

Work Branch

Use branch format:

deepwork/manual_tests-[instance]-YYYYMMDD

• If on a matching work branch: continue using it
• If on main/master: create new branch with

  git checkout -b deepwork/manual_tests-[instance]-$(date +%Y%m%d)

Outputs

Required outputs:

• fire_results

Guardrails

• Do NOT skip prerequisite verification if this step has dependencies
• Do NOT produce partial outputs; complete all required outputs before finishing
• Do NOT proceed without required inputs; ask the user if any are missing
• Do NOT modify files outside the scope of this step's defined outputs

Quality Validation

Before completing this step, you MUST have your work reviewed against the quality criteria below.

Use a sub-agent (Haiku model) to review your work against these criteria:

Criteria (all must be satisfied):

1. Sub-Agents Used: Did the main agent spawn a sub-agent (using the Task tool) for EACH test? The main agent must NOT edit the test files directly.
2. Sub-Agent Config: Did all sub-agents use

   model: "haiku"

   and

   max_turns: 5

   ?
3. Serial Execution: Were sub-agents launched ONE AT A TIME (not in parallel) to prevent cross-contamination?
4. Hooks Fired Automatically: Did the main agent observe the blocking hooks firing automatically when each sub-agent returned? The agent must NOT manually run the rules_check command.
5. Reset Between Tests: Was the reset step called internally after each test to revert files and prevent cross-contamination?
6. Early Termination: If 2 tests failed, did testing halt immediately with results reported?
7. Results Recorded: Did the main agent track pass/fail status for each test case? Review Process:
8. Once you believe your work is complete, spawn a sub-agent using Haiku to review your work against the quality criteria above
9. The sub-agent should examine your outputs and verify each criterion is met
10. If the sub-agent identifies valid issues, fix them
11. Have the sub-agent review again until all valid feedback has been addressed
12. Only mark the step complete when the sub-agent confirms all criteria are satisfied

On Completion

1. Verify outputs are created
2. Inform user: "run_all step 3/4 complete, outputs: fire_results"
3. Continue workflow: Use Skill tool to invoke

   /manual_tests.infinite_block_tests

---

Reference files:

.deepwork/jobs/manual_tests/job.yml

,

.deepwork/jobs/manual_tests/steps/run_fire_tests.md