Reference index

Progressive Harsh Review

Mechanical routing: Run tools/review.sh route <path> [<path> ...] first. Obey its result; stop on error or an unclassified artifact.

Wrong skill? Code -> code-review-battery; skills/tooling -> llm-skill-review; design comparison -> debate.

Announce at start: "I'm using the progressive-harsh-review skill to red-team this work."

When to use

Use before presenting non-code work or on a hostile-review request. Exclude code, skills, brainstorming, and design-option selection.

Persona dimension table

Persona count: all three for irreversible artifacts (ADRs, API/contract specs, migrations, security policy, author-marked); otherwise SeniorArchCritic alone; unsure -> three. Each persona reads independently. Send only its row, the common dimension questions, and artifact/repository access.

Dimensions: Correctness (holds?), Simplicity (needless complexity?), Verifiability (checkable?), Blind Spots (omissions?), Operational Risk (adverse failures?). Each row totals 100.

For user-visible functionality, missing named metrics and trace/span strategy caps OpsRealist's Operational Risk at 4; cite the omission.

Review process

1. Fresh-reader check. Flag local paths, undefined identifiers, process commentary, and inaccessible references. Remove this author noise before shipping; do not lower scores for it alone.
2. Independent review. Author != Reviewer. Dispatch the personas chosen above from artifact paths with only their row, five questions, and repository access. Persona reviewers must not invoke PHR, debate, code-review-battery, or other reviewers; each returns one scorecard. Remediate only after aggregation.
3. Score. Score each dimension 1-10 with the persona's weights; with three, take the equal-weight average of the weighted scores.
4. Apply veto. Correctness or Operational Risk <=4 is a hard veto only with a specific defect. Unrecoverable failures must affect Operational Risk, not Blind Spots alone.
5. Verdict. Use the table. A repository floor raises the PASS bar only; load Project-min override.
6. Remediate. Fix and verify every material finding. After round one, continue the same reviewer on the delta (a fresh one needs the full prior findings). REJECT requires root-cause analysis and full re-review.
7. Check correlation. Any flag below or unsupported clean sweep requires a new persona starting point.
8. Converge. Require the active floor, no veto/flag, and no new material issues. Escalate after 3 rounds without convergence; never auto-ship.

Verdicts

PASS_WITH_FIXES never clears the gate. Limit the entire remediation cycle to 3 review rounds.

Correlated-failure checks

• CORRELATED EVIDENCE: shared evidence; one persona restarts from its lens.
• ECHO REASONING: materially identical reasoning; require an independent restatement.
• Clean sweep: each persona shows evidence from its distinct start point or re-examines.

Sentinel after PASS

Only PASS clears the gate. Run PHR AFTER git commit once the floor is met with no veto or correlation flag:

tools/run-phr.sh --verdict PASS --min-score "<weighted-mean>"

Clearance holds only while the reviewed docs are byte-identical (tools/lib/sentinel-scope.sh).

Reference loading

Load only the needed section: Project-min override, Anti-Patterns, or Scoring output format (see the routing table above for exact heading names -- the loader below requires an exact match). When an installed copy is found (any of .claude/.codex/.agents), loading requires the separately-provisioned ~/.codex/superpowers-plus checkout -- an install missing that shared dependency gets a loud section-loader missing failure, not silent wrong content.

_ks_ref=""
_ks_loader=""
for _candidate in \
  "$HOME/.claude/skills/sp-phr/reference.md" \
  "$HOME/.codex/skills/sp-phr/reference.md" \
  "$HOME/.agents/skills/sp-phr/reference.md"
do
  if [ -r "$_candidate" ]; then _ks_ref="$_candidate"; break; fi
done
if [ -n "$_ks_ref" ]; then
  _ks_loader="$HOME/.codex/superpowers-plus/tools/section-loader.sh"
else
  _project_root="$(git rev-parse --show-toplevel 2>/dev/null || true)"
  _source_dir="$_project_root/skills/engineering/progressive-harsh-review"
  if [ -n "$_project_root" ] && [ -r "$_source_dir/skill.md" ] && \
     [ -r "$_source_dir/reference.md" ] && \
     [ -r "$_project_root/tools/section-loader.sh" ]; then
    _ks_ref="$_source_dir/reference.md"
    _ks_loader="$_project_root/tools/section-loader.sh"
  fi
fi
[ -r "$_ks_ref" ] || { printf 'reference missing\n' >&2; exit 1; }
[ -r "$_ks_loader" ] || { printf 'section-loader missing\n' >&2; exit 1; }
# Replace <section heading> below with one of the exact strings from
# the Reference index table above before running.
_section='<section heading>'
bash "$_ks_loader" "$_ks_ref" "$_section" \
  || { printf 'section not found: %s\n' "$_section" >&2; exit 1; }

Failure Modes