judge

Research-supervisor review of program.md — validates experimental methodology, emits APPROVED / NEEDS-REVISION / BLOCKED verdict before expensive run loop. Read-only; never modifies code or state.

NOT for: running experiments (use /research:run); designing hypotheses (use research:scientist agent); config quality (/foundry:audit (requires foundry plugin)).

• Key boundary: end of J3 — methodology and scientific review agents complete, output files written; before J4 validation and J6 verdict.
• Preserve: RUN_DIR (TMPDIR key), PROGRAM_PATH (TMPDIR key), methodology.md path, scientific-review.md path, SKIP_VALIDATION flag, BRANCH.
• Clear at J1 start (stale prior run) and at end of J6 after terminal summary printed.

Agent Resolution

Agent resolution: load and follow the protocol below. Contains foundry check + fallback table. If foundry not installed: use table to substitute each foundry:X with general-purpose. Agents: foundry:solution-architect, research:scientist.

# loads: compaction-contract.md
export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
_RESEARCH_SHARED=$(python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/resolve_shared.py" 2>/dev/null)  # timeout: 5000
[ -z "$_RESEARCH_SHARED" ] && { echo "! Plugin path resolution failed — ensure research plugin installed and CLAUDE_PLUGIN_ROOT set, or invoke from project root."; exit 1; }
echo "$_RESEARCH_SHARED" > "${TMPDIR:-/tmp}/research-shared-${CSID}"  # cold resolve — every later site reads this sentinel instead of re-running python
cat "$_RESEARCH_SHARED/agent-resolution.md"

Agent	Fallback if absent
foundry:solution-architect	general-purpose (methodology review quality reduced — ⚠ general-purpose agent may not emit methodology_rating in required format; verdict defaults to NEEDS-REVISION)
research:scientist	general-purpose (scientific rigor review quality reduced — ⚠ general-purpose agent may not emit scientific_rating; verdict defaults to NEEDS-REVISION)

Judge Mode (Steps J1–J6)

Triggered by judge or judge <file.md>.

Task tracking: create tasks for J1, J2, J3, J4, J5a, J5b, J6 at start — before any tool calls. (J5a = Codex adversarial review; J5b = resolve rating source.)

Step J1: Locate and parse program.md

Flag parsing (first):

SKIP_VALIDATION=false
[[ "$ARGUMENTS" == *"--skip-validation"* ]] && SKIP_VALIDATION=true
ARGUMENTS="${ARGUMENTS/--skip-validation/}"
ARGUMENTS="${ARGUMENTS#"${ARGUMENTS%%[![:space:]]*}"}"  # trim leading whitespace

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/extract-keep-flag.py" judge "$ARGUMENTS"  # timeout: 5000 — parses --keep, clears a stale contract, persists for J3

Unsupported flag check: load and follow the protocol below. Supported flags for this skill: --skip-validation, --keep.

# loads: unsupported-flag-protocol.md
export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
IFS= read -r _RESEARCH_SHARED < "${TMPDIR:-/tmp}/research-shared-${CSID}" 2>/dev/null || _RESEARCH_SHARED=""  # warm read (Check 41)
cat "$_RESEARCH_SHARED/unsupported-flag-protocol.md"

Input resolution (priority order):

1. Explicit argument: /research:judge path/to/plan.md
2. Auto-detect: program.md at project root
3. Latest state: scan .experiments/state/*/state.json for most recent with status: running and non-null program_file field
4. If nothing found: stop with error:

   No program.md found. Run /research:plan <goal> first, or provide a path: /research:judge <path.md>
   

Parsing — find ## <Section> headings in program.md, extract first fenced code block per section, parse as key: value lines, warn on unrecognized keys. --skip-validation and colab_hw judge-specific, extracted independently.

Placeholder substitution — after parsing, apply same substitution as R1: resolve all {field_name} tokens in metric_cmd and guard_cmd using ## Config fields, fallback to declared default. No clarification_prompt in judge — skip clarification-override step.

Extract <program_title> from # Program: <title> line for reports (fallback # Campaign: <title> for legacy files).

Step J2: Completeness audit

Check C1–C12 plus C2b and C6b (14 items total). Produce findings list with severity. Each finding has: id, check, status (pass/fail/warn), severity, detail.

ID	Check	Severity if failing	Description
C1	## Goal present and non-empty	critical	Campaign cannot run without a goal
C2	## Metric has command field	critical	No metric = no feedback loop
C3	## Metric has direction field (higher/lower)	critical	Cannot decide keep/revert without direction
C4	## Guard has command field	critical	Without guard, regressions go undetected. Note: a command field containing only echo 0, true, or exit 0 is equivalent to no guard (always exits 0 regardless of test state) — flag as critical with detail "guard command is a no-op; add real regression detection".
C5	scope_files present in ## Config	high	Without scope, ideation agent modifies arbitrary files
C6	Each scope_files path exists on disk (glob match)	high	Non-matching patterns = ideation agent has nothing to work with. If filesystem unavailable, flag warn unless path name signals non-existence (e.g., nonexistent, placeholder, todo, legacy_v1, deprecated, old, removed).
C7	target set in ## Metric	medium	Without target, campaign runs to max_iterations — may waste compute
C8	max_iterations in bounds (1–50)	medium	Missing defaults to 20 (acceptable); >50 violates SKILL.md constants. Additionally: if value is within bounds but >20 AND combined with risk factors (C4 fails / guard empty, OR C6 fails / scope non-existent), add a separate low finding: "max_iterations=N is elevated; with no functioning guard/scope, runaway iterations amplify risk — consider reducing to ≤15 until guard/scope is fixed"
C9	agent_strategy is valid (auto/perf/code/ml/arch)	medium	Invalid value silently falls back to auto
C10	compute is valid (local/colab/docker)	low	Invalid defaults to local
C11	colab_hw valid (if present)	low	colab_hw absent OR is one of H100, L4, T4, A100 — fail detail: "colab_hw '<value>' is not in known set {H100, L4, T4, A100} — run mode will warn and null this value to the default GPU". Note: this check is a minimum-capability floor — new Colab hardware tiers may exist beyond this list; unknown values are flagged for user verification, not blocked.
C12	## Notes section present	low	Notes optional but improve ideation quality

Scope adequacy sub-rule (C6b) — after C6 passes, assess whether scope_files is sufficient for stated goal. If goal type implies known bottleneck locations outside declared scope, add medium finding:

• Test-speed goal + scope limited to tests/ only → flag: "conftest.py, fixtures, and test infrastructure outside tests/ are common levers for test runtime; scope may be too narrow"
• Throughput/latency goal + scope limited to single-layer path (e.g., src/serving/) → flag: "serving bottlenecks often span middleware, connection pooling, or database layers outside declared scope"
• Any goal where the stated scope excludes a widely-known dependency class → emit medium finding with location ## Config / scope_files, suggested broader pattern as fix

Distinct from C6 (path existence) — C6b fires even when path exists but is likely insufficient.

Severity summary: count findings per severity. Any critical finding = verdict cannot be APPROVED. Enumeration rule: check ALL items (C1–C12 plus C2b and C6b) before stopping — don't short-circuit after first critical issue. program.md can have multiple independent flaws across severity levels; Required Changes section must list all, not just verdict-determining one.

Placeholder token check (C2, C4 sub-rule) — after confirming command present in ## Metric (C2) and ## Guard (C4), scan each command for {...} tokens. Verify each token's field name exists in ## Config. Token with no matching field = unresolvable — add high finding. Don't flag {field_name} tokens as malformed; valid when resolvable.

Goodhart's Law check (C2b) — after confirming metric command present (C2 passes), assess whether command operationalizes stated ## Goal or measures proxy. If metric could improve while actual goal NOT achieved, add critical finding:

• metric measures test pass rate but goal is latency reduction → critical: "metric is a correctness proxy, not a latency measure"
• metric measures lint error count but goal is bug density reduction → critical: "pylint score is a gameable proxy; agent can suppress warnings without improving actual quality"
• metric measures a format/style score but goal is functional improvement → critical: "metric does not operationalize the stated goal"

Goodhart findings are critical (not just methodology notes) — broken metric invalidates entire feedback loop, equivalent impact to C2 (missing command).

Command feasibility: J2 validates command fields statically (presence, format). Executability deferred to J4. If $SKIP_VALIDATION is true, J4 skipped, commands unverified — report as "validation skipped — commands unverified."

Step J3: Methodology review

Pre-compute run dir before spawning:

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
RUN_DIR=$(python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/make_run_dir.py" "judge" ".experiments" 2>/dev/null)  # timeout: 5000
echo "$RUN_DIR" > "${TMPDIR:-/tmp}/judge-run-dir-${CSID}"  # persist for J3 block (Check 41)

Synchronous spawn note: J3 agents spawned synchronously (not run_in_background=true), so CLAUDE.md §6 sentinel polling unreachable mid-call. Timeout handled post-hoc — after each Agent() returns, check output file; if missing/empty mark agent timed out (⏱). See J3 post-call checks below.

Dispatch — scientist dimension always; architect dimension only when the complexity gate fires. When BOTH dimensions are active, J3 is ONE merged spawn covering both (two opus spawns for a single-file review pay 2× ~120,851 tok of fixed overhead for a workload far under the breakeven above); when only the scientist dimension is active, it is a single scientist spawn as before. Per-dimension output files and Confidence blocks are unchanged either way.

Before constructing J3 prompts, expand all bash variables into concrete paths — never pass literal <path_to_program.md> or <RUN_DIR> placeholders to agents:

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
PROGRAM_PATH=$(realpath "$PROGRAM_FILE" 2>/dev/null || echo "$PROGRAM_FILE")
echo "$PROGRAM_PATH" > "${TMPDIR:-/tmp}/judge-program-path-${CSID}"  # persist for J3 complexity-gate block (Check 41)
# Reload RUN_DIR (Check 41: fresh shell per call — persisted in J2 block)
IFS= read -r RUN_DIR < "${TMPDIR:-/tmp}/judge-run-dir-${CSID}" 2>/dev/null || RUN_DIR=""

Compute SKIP_VALIDATION_NOTE before constructing the prompt:

if [ "${SKIP_VALIDATION:-false}" = "true" ]; then
    SKIP_VALIDATION_NOTE="Local validation skipped via --skip-validation — do NOT assess executability of metric_cmd/guard_cmd; note this limitation in your review."
else
    SKIP_VALIDATION_NOTE="Local validation will run after this review (J4)."
fi

Complexity gate — mirrors P-P2b; skip architect for narrow single-scope experiments (saves full opus pass):

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
IFS= read -r PROGRAM_PATH < "${TMPDIR:-/tmp}/judge-program-path-${CSID}" 2>/dev/null || PROGRAM_PATH=""  # re-hydrate (Check 41: fresh shell — persisted in J3 pre-spawn block)
SPAWN_ARCHITECT=$(python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/detect-complexity.py" "$PROGRAM_PATH")  # timeout: 5000

When SPAWN_ARCHITECT=false: skip architect spawn; J5b precedence step 0 sets methodology_rating="sound" (scientist review still covers scientific rigor); record architect: skipped (narrow scope) in J6 summary.

Agent budget — each spawn costs ~120,851 tok of fixed overhead (~73 tool-calls' worth) plus ~12.0 s/call, so work under ~73 calls is cheaper done inline: spawn nothing. Keep each agent near ~55 tool-calls; past ~60 they stall without returning an envelope, forcing reconstruction from disk. Every spawn prompt must require an envelope even on exhaustion — partial: true plus what was finished.

When SPAWN_ARCHITECT=true: spawn ONE merged reviewer via Agent(subagent_type="research:scientist", prompt=<merged prompt>) (opus) — build the merged prompt exactly per judge-j3-prompts.md §Merged spawn — envelope override (both templates with their individual Return ONLY: lines stripped, joined by ---, closed with the override's single array-envelope instruction). The single spawn executes BOTH templates in order and writes ${RUN_DIR}/methodology.md AND ${RUN_DIR}/scientific-review.md, each with its own Confidence block. Full prompt templates (expand ${PROGRAM_PATH} and ${RUN_DIR} before passing):

$J3_ARCH_PROMPT template externalized — load and follow protocol below § J3_ARCH_PROMPT (one load supplies both J3 templates).

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
IFS= read -r _RESEARCH_SHARED < "${TMPDIR:-/tmp}/research-shared-${CSID}" 2>/dev/null || _RESEARCH_SHARED=""  # warm read (Check 41)
cat "$_RESEARCH_SHARED/judge-j3-prompts.md"

Substitution requirement: every ${RUN_DIR} and ${PROGRAM_PATH} token in template above MUST be replaced with concrete bash-expanded value (e.g. .experiments/judge-2026-05-13T10-00-00Z) before string passed to Agent(...). Passing literal ${RUN_DIR} to agent causes agent to write to directory named ${RUN_DIR}. Applies equally to any historical <RUN_DIR> angle-bracket notation in older copies — both forms are text-substitution placeholders, not bash interpolation the Agent runtime expands.

When SPAWN_ARCHITECT=true — after the merged Agent() returns, check $RUN_DIR/methodology.md: if missing or empty, set methodology_rating = "timed_out", continue to J6; surface with ⏱ in report.

After the spawn returns (merged or scientist-only), check $RUN_DIR/scientific-review.md: if missing or empty, set scientific_rating = "timed_out", continue to J6; surface with ⏱ in Scientific Rigor section.

Use methodology_rating from returned envelope for verdict computation in J6:

• sound → supports APPROVED
• needs-refinement → supports NEEDS-REVISION
• fundamentally-flawed → supports BLOCKED

When SPAWN_ARCHITECT=false: spawn research:scientist alone for scientific rigor. Expand ${PROGRAM_PATH} and ${RUN_DIR} before passing — construct $J3_SCI_PROMPT per the template (same variable substitution pattern). Spawn: Agent(subagent_type="research:scientist", prompt=$J3_SCI_PROMPT). (When SPAWN_ARCHITECT=true the scientist dimension already ran inside the merged spawn above — do NOT spawn a second agent.)

$J3_SCI_PROMPT template: judge-j3-prompts.md § J3_SCI_PROMPT (already loaded by J3_ARCH cat above).

Use scientific_rating as advisory in J6 report under Scientific Rigor — informs but doesn't override verdict. Exception: scientific_rating == "fundamentally-flawed" (exact match) elevates verdict to BLOCKED.

Source precedence for scientific_rating (mandatory when both present):

1. File-parsed value from $RUN_DIR/scientific-review.md (read after agent completes) — authoritative
2. Health-monitor / envelope value from agent's returned JSON — advisory only

File-parsed value takes priority over health monitor value; use file-parsed value when both present. Same precedence applies to methodology_rating parsed from $RUN_DIR/methodology.md vs envelope value. Use envelope value only when file missing or unparsable (e.g., timeout with no output).

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
# boundary: after J3 agents complete, before J4 validation (compaction-contract.md)
IFS= read -r _RUN_DIR < "${TMPDIR:-/tmp}/judge-run-dir-${CSID}" 2>/dev/null || _RUN_DIR=""
IFS= read -r _PROG_PATH < "${TMPDIR:-/tmp}/judge-program-path-${CSID}" 2>/dev/null || _PROG_PATH=""
IFS= read -r _KEEP < "${TMPDIR:-/tmp}/judge-keep-items-${CSID}" 2>/dev/null || _KEEP=""
_KEEP_APPEND=""; [ -n "$_KEEP" ] && _KEEP_APPEND="; user-keep: $_KEEP"
python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/write-skill-contract.py" "research:judge" "validation-verdict (after J3 review agents complete)" "${_RUN_DIR}" "run-dir=${_RUN_DIR}, program=${_PROG_PATH}, methodology=${_RUN_DIR}/methodology.md, scientific-review=${_RUN_DIR}/scientific-review.md${_KEEP_APPEND}" "J4 local validation → J5 Codex review → J6 verdict and report"  # timeout: 5000

Step J4: Local validation

Skip if $SKIP_VALIDATION is true (parsed in J1). Print: → Validation skipped (--skip-validation passed) and continue to J5. Add high finding: "Executability unverified — metric_cmd and guard_cmd not tested on local machine." This finding persists into J6 — APPROVED not achievable when --skip-validation set (high > 0 → NEEDS-REVISION at best).

Execute each command once. Non-blocking — failures become critical findings, not hard stops.

Substitution invariant — metric_cmd and guard_cmd fully resolved in J1. No {...} tokens should remain. If any {field_name} token still present, add critical finding: "Unresolved placeholder {field_name} in <metric_cmd|guard_cmd> — substitution failed in J1" and skip execution.

# Substitute ${metric_cmd} with the resolved command from J1 before execution
${metric_cmd} 2>&1  # timeout: 360000

Parse stdout for float. If found, record as baseline_value. If not found or non-zero exit: add critical finding: "Metric command failed or produced no numeric output".

# Substitute ${guard_cmd} with the resolved command from J1 before execution
${guard_cmd}  # timeout: 360000

If guard exits non-zero: add critical finding: "Guard command exited non-zero (exit ): <first 3 lines of output>".

Record validation results for J6 report.

Note: J4 executes on current machine. For cross-machine workflows, pass --skip-validation.

Step J5a: Codex adversarial review

Complexity gate — simple programs skip the Codex pass (J2 completeness, scientist review, and J4 dry-run still apply):

export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
IFS= read -r PROGRAM_PATH < "${TMPDIR:-/tmp}/judge-program-path-${CSID}" 2>/dev/null || PROGRAM_PATH=""  # re-hydrate (Check 41: fresh shell)
J5A_COMPLEX=$(python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/detect-complexity.py" "$PROGRAM_PATH")  # timeout: 5000 — same gate as J3; must stay identical, hence the shared script

J5A_COMPLEX=false → print note: simple program (single scope file, single-phase strategy) — Codex adversarial pass skipped by complexity gate, record codex: skipped (complexity gate) in J6 summary, continue to J5b. J5A_COMPLEX=true → proceed below.

Check Codex availability. Distinguish two failure modes — CLI missing vs plugin missing:

if ! command -v claude >/dev/null 2>&1; then
    CODEX_STATUS="cli-missing"
elif [ "$(python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_research}/bin/check_bridge.py" --status 2>/dev/null || echo absent)" = "available" ]; then
    CODEX_STATUS="available"
else
    CODEX_STATUS="plugin-missing"
fi

CODEX_STATUS=available: invoke adversarial review on top 3 critical/high gaps from J2 and J3. Example (replace <top finding N> with actual findings):

Skill(skill="bridge:review", args="Read-only adversarial review of run program. Check <top finding 1>, <top finding 2>, and <top finding 3> in program.md; do not apply fixes.")

Incorporate Codex findings into overall findings list with source: "codex".

CODEX_STATUS=plugin-missing: print one line and continue:

note: bridge@borda-ai-rig not installed or disabled — skipping adversarial review (Claude-only judge)

CODEX_STATUS=cli-missing: print diagnostic and continue (distinguish from plugin-absent so user with Codex installed but claude CLI not in PATH isn't silently denied review):

note: `claude` CLI not in PATH — Codex availability cannot be verified; skipping adversarial review. To enable: ensure `claude` binary is on PATH and Codex plugin installed.

Step J5b: Resolve rating source

Apply rating source precedence before J6 verdict computation — fixes ambiguity when envelope and file-parsed ratings disagree. For methodology_rating:

0. If SPAWN_ARCHITECT=false (complexity gate didn't fire): set methodology_rating="sound" directly — no file or envelope; log → methodology_rating: sound (architect gate skipped — narrow scope). Skip steps 1–2 for methodology_rating.
1. If $RUN_DIR/methodology.md present AND parsable, use file-parsed value — authoritative.
2. Else use envelope value from agent's returned JSON — fallback.
3. Log source used: print → methodology_rating source: file | envelope.

For scientific_rating:

1. If $RUN_DIR/scientific-review.md present AND parsable, use file-parsed value — authoritative.
2. Else use envelope value from the agent's returned JSON — fallback.
3. Log source used: print → scientific_rating source: file | envelope.

Resolved rating values feed directly into J6 verdict table — no further source disambiguation in J6.

Step J6: Verdict and report

Verdict computation (deterministic — design soundness, not outcome prediction):

Top-to-bottom; first match wins. BLOCKED takes precedence — stop at first match.

Condition	Verdict
any critical (J2) — exact critical severity match	BLOCKED
methodology_rating == "fundamentally-flawed" (exact string match, J3)	BLOCKED
scientific_rating == "fundamentally-flawed" (exact string match, J3)	BLOCKED
J3 agent timed out (methodology_rating == "timed_out" — exact match — or null; note: timed_out does not trigger BLOCKED — it falls to NEEDS-REVISION)	NEEDS-REVISION
scientific_rating == "timed_out" (exact match — scientist review did not complete; adversarial review absent → APPROVED not safe)	NEEDS-REVISION
0 critical AND (high > 0 OR methodology_rating == "needs-refinement")	NEEDS-REVISION
0 critical AND 0 high AND methodology_rating == "sound" AND scientific_rating != "timed_out"	APPROVED

Verdict matching rules: all *_rating comparisons require exact string match. Reject partial/substring matches — e.g., timed_out_partial does NOT match timed_out; flawed does NOT match fundamentally-flawed. Use == equality only; never =~, startswith, or pattern matching.