pj f933ce6d15 Record working files written outside the scratch directory ...

Makes a Phase 2 null result interpretable rather than ambiguous. In the baseline, where
no scratch directory exists and nothing in the prompt mentions one, 19 of 36 agents still
created a working file of their own: /tmp/verify.py, /tmp/brute.py, /tmp/proto.py, and in
two cases scratch.py beside their work.

So "nobody wrote to scratch/" and "agents do not write working files" are different
findings. Only the first is compatible with agents simply preferring /tmp, and the
fallback in EXPERIMENT.md is the right response to one and not the other. Without this
signal the pilot cannot tell them apart.

Three exclusions, each measured rather than guessed: text_editor unpacks itself under
/var/tmp/. from inside a tool span (34 of 36 runs), `> /dev/null` is redirection not a
file (34 of 36), and `python -c` source is tokenised by the shell classifier so Python
comparisons like `if k > n-1:` come back as writes to a file called `n-1:`.

Claude-Session: https://claude.ai/code/session_01Cq98H7sNoSJdL3W98f18bu

2026-08-31 22:38:48 +05:30

..

test_analysis.py

Do not attribute the harness's own filesystem actions to the agent

2026-08-31 21:43:22 +05:30

test_byte_match.py

Scratch directory task, event adapter, per-run checks

2026-08-31 21:36:13 +05:30

test_checks.py

Record working files written outside the scratch directory

2026-08-31 22:38:48 +05:30

test_events.py

A file called scratch.py is not the scratch directory

2026-08-31 22:02:56 +05:30

test_prompt_parity.py

Guard that the fork presents the same task text as the stock benchmark

2026-08-31 21:40:09 +05:30

test_shell.py

Scratch directory task, event adapter, per-run checks

2026-08-31 21:36:13 +05:30