agentv 4.26.1 → 4.27.0-next.1

package/dist/skills/agentv-bench/SKILL.md

@@ -0,0 +1,459 @@
+---
+name: agentv-bench
+description: >-
+  Run AgentV evaluations and optimize agents through eval-driven iteration.
+  Triggers: run evals, benchmark agents, optimize prompts/skills against evals, compare
+  agent outputs across providers, analyze eval results, offline evaluation of recorded sessions,
+  run autoresearch, optimize unattended, run overnight optimization loop.
+  Not for: writing/editing eval YAML without running (use agentv-eval-writer),
+  analyzing existing traces/JSONL without re-running (use agentv-trace-analyst).
+---
+
+# AgentV Bench
+
+
+A skill for evaluating agents and iteratively improving them through data-driven optimization.
+
+At a high level, the process goes like this:
+
+- Understand what the agent does and what "good" looks like
+- Write evaluation test cases (EVAL.yaml or evals.json)
+- Run the agent on those test cases, grade the outputs
+- Analyze the results — what's working, what's failing, and why
+- Improve the agent's prompts/skills/config based on the analysis
+- Repeat until you're satisfied
+
+Your job when using this skill is to figure out where the user is in this process and then jump in and help them progress. Maybe they want to start from scratch — help them write evals, run them, and iterate. Maybe they already have results — jump straight to analysis and improvement.
+
+Be flexible. If the user says "I don't need a full benchmark, just help me debug this failure", do that instead.
+
+After the agent is working well, you can also run description optimization to improve skill triggering accuracy (see `references/description-optimization.md`).
+
+## Communicating with the user
+
+This skill is used by people across a wide range of familiarity with evaluation tooling. Pay attention to context cues:
+
+- "evaluation" and "benchmark" are borderline but OK in most cases
+- For "YAML", "grader", "assertion", "deterministic judge" — see serious cues from the user that they know what those mean before using them without explanation
+- Briefly explain terms if in doubt
+
+When presenting results, default to summary tables. Offer detail on request. In CI/headless mode, skip interactive prompts and exit with status codes.
+
+---
+
+## Step 1: Understand the Agent
+
+Before running or optimizing, understand what you're working with.
+
+1. **Read the agent's artifacts** — prompts, skills, configs, recent changes. Understand the full picture: what tools are available, what the expected input/output looks like, what constraints exist.
+
+2. **Identify success criteria** — what does "good" look like for this agent? What are the edge cases? What would a failure look like? Talk to the user if this isn't clear from the artifacts alone.
+
+3. **Understand the target harness** — which provider runs the agent (Claude, GPT, Copilot CLI, Gemini, custom CLI)? This affects what grader types are available and how to run tests. Targets are configured in `.agentv/targets.yaml` (canonical location, searched from the eval file directory upward). Sensitive values like `api_key` must use `${{ ENV_VAR }}` syntax — literal secrets are rejected as a security guardrail.
+
+4. **Challenge assumptions** — if evals already exist, review their quality before running:
+   - Are the test cases testing the right things?
+   - Are assertions specific enough to catch real failures?
+   - Are there ambiguous or contradictory test cases?
+   - Flag eval issues before proceeding — running bad evals wastes time.
+
+5. **Check integrity** — ensure task prompts (what the agent receives) are not also used as grader prompts (how outputs are scored). If a prompt file appears in both locations, note the overlap and optimize only for the task purpose.
+
+---
+
+## Step 2: Write Evaluations
+
+AgentV supports two evaluation formats:
+
+**EVAL.yaml** (native, full features) — supports workspaces, code graders, multi-turn conversations, tool trajectory scoring, workspace file tracking, multi-provider targets. Use this for agent evaluation.
+
+```yaml
+# example.eval.yaml
+tests:
+  - id: basic-code-review
+    input: "Review this TypeScript file for bugs and suggest improvements"
+    criteria: "Identifies the null pointer bug on line 12 and suggests a fix"
+    assertions:
+      - type: contains
+        value: "null"
+      - Review identifies the null pointer bug and suggests a concrete fix
+
+workspace:
+  template: ./workspace-template
+  hooks:
+    before_each:
+      reset: fast
+```
+
+Multi-skill evaluation is handled naturally via input messages — describe the task in the test input, and the agent uses whatever skills it needs.
+
+**evals.json** (skill-creator compatible) — auto-promoted to EVAL-equivalent format:
+- `prompt` → input messages
+- `expected_output` → reference answer
+- `assertions` → graders
+- `files[]` paths resolved relative to the evals.json location
+
+```json
+{
+  "skill_name": "my-agent",
+  "evals": [
+    {
+      "id": 1,
+      "prompt": "User's task prompt",
+      "expected_output": "Description of expected result",
+      "assertions": ["Output includes error handling", "Uses async/await"]
+    }
+  ]
+}
+```
+
+### Writing good test cases
+
+Start with 2-3 realistic test cases — the kind of thing a real user would actually say. Share them with the user before running: "Here are a few test cases I'd like to try. Do these look right, or do you want to add more?"
+
+Good assertions are objectively verifiable and have descriptive names. Subjective quality ("the output is good") is better evaluated qualitatively — don't force assertions onto things that need human judgment.
+
+**Grader types** (cheapest to most expensive): `exact`, `contains`, `regex`, `is-json`, `field-accuracy`, `composite`, `code-grader`, `tool-trajectory`, `llm-grader`. See `references/eval-yaml-spec.md` for full config and grading recipes for each type.
+
+Prefer deterministic graders over LLM graders whenever possible. If an assertion can be checked with `contains` or `regex`, don't use `llm-grader`.
+
+---
+
+## Step 3: Run and Grade
+
+This section is one continuous sequence — don't stop partway through.
+
+Each run produces a new `.agentv/results/runs/<timestamp>/` directory automatically. Use timestamps to identify iterations when comparing runs.
+
+### Choosing a run mode
+
+**User instruction takes priority.** If the user says "run in subagent mode", "use subagent mode", or "use CLI mode", use that mode directly.
+
+If the user has not specified a mode, default to `subagent`.
+
+| `AGENT_EVAL_MODE` | Mode | How |
+|----------------------|------|-----|
+| `subagent` (default) | **Subagent mode** | Subagent-driven eval — parses eval.yaml, spawns executor + grader subagents. Zero CLI dependency. |
+| `cli` | **AgentV CLI** | `agentv eval <path>` — end-to-end, multi-provider |
+
+Set `AGENT_EVAL_MODE` in `.env` at the project root as the default when no mode is specified. If absent, default to `subagent`. **User instruction always overrides this.**
+
+**`subagent`** — Parses eval.yaml directly, spawns executor subagents to run each test case in the current workspace, then spawns grader subagents to evaluate all assertion types natively. No CLI or external API calls required. Read `references/subagent-pipeline.md` for the detailed procedure.
+
+**`cli`** — AgentV CLI handles execution, grading, and artifact generation end-to-end. Works with all providers. Use when you need multi-provider benchmarking or CLI-specific features.
+
+### Running evaluations
+
+**AgentV CLI mode** (end-to-end, EVAL.yaml):
+```bash
+agentv eval <eval-path> --output .agentv/artifacts/
+```
+
+**Subagent mode** — read `references/subagent-pipeline.md` for the detailed procedure. In brief: use `pipeline input` to extract inputs, dispatch one `executor` subagent per test case (all in parallel), then proceed to grading below.
+
+**Spawn all runs in the same turn.** For each test case that needs both a "with change" and a "baseline" run, launch them simultaneously. Don't run one set first and come back for the other — launch everything at once so results arrive around the same time.
+
+**Multi-target benchmarking:**
+```bash
+agentv eval <eval-path> --target claude --target gpt --target copilot
+```
+
+**Baseline strategy:**
+- **New agent**: baseline is "no prompt" or minimal prompt — same eval, no agent-specific configuration
+- **Improving existing**: snapshot the current version before editing (`cp -r <prompt-dir> <workspace>/prompt-snapshot/`), use as baseline throughout
+- **Multi-target**: each target is its own baseline — no need for a separate "without" run
+
+### While runs are in progress, draft graders
+
+Don't just wait for runs to finish — use this time productively. If assertions don't exist yet, draft them now. If they exist, review them and explain what they check to the user.
+
+Good assertions are *discriminating* — they pass when the agent genuinely succeeds and fail when it doesn't. An assertion that passes for both good and bad outputs is worse than no assertion.
+
+### As runs complete, capture timing data
+
+When each subagent task completes, you receive a notification containing `total_tokens` and `duration_ms`. **Save this data immediately** to `timing.json` in the run directory. See `references/schemas.md` for the timing.json schema.
+
+This is the only opportunity to capture this data — it comes through the task notification and isn't persisted elsewhere. Process each notification as it arrives.
+
+### After all executors complete
+
+Once all executor subagent notifications arrive, **read `response.md` directly from disk**.
+Do NOT use `read_agent` to fetch executor results — they are already written to the run
+directory. Verify all responses exist before proceeding:
+
+```bash
+for d in <run-dir>/<evalset>/*/; do
+  echo "$(basename $d): $(ls "$d"/response.md 2>/dev/null && echo OK || echo MISSING)"
+done
+```
+
+If any `response.md` is MISSING, re-run that specific executor subagent. Do not proceed to
+grading until all responses are present.
+
+### Grading
+
+**In CLI mode**, `agentv eval` handles all grading end-to-end — no manual phases needed.
+
+**In subagent mode**, grading has three phases. **All three are required — do not stop after phase 1.**
+
+**Phase 1: Code graders** (deterministic, zero-cost)
+
+```bash
+agentv pipeline grade <run-dir>
+```
+
+This evaluates all deterministic assertions against `response.md` files. Two types are handled:
+- **`code-grader` scripts** — external scripts executed against the response (arbitrary logic, any language)
+- **Built-in assertion types** — evaluated in-process: `contains`, `contains-any`, `contains-all`, `icontains`, `regex`, `equals`, `starts-with`, `ends-with`, `is-json`, and variants
+
+Both types are configured by `pipeline input` into `code_graders/<name>.json` and graded by `pipeline grade`. Results are written to `<test-id>/code_grader_results/<name>.json`. Alternatively, pass `--grader-type code` to `pipeline run` to run these inline.
+
+**Do not dispatch LLM grader subagents for tests that only have `contains`, `regex`, or other built-in assertions** — `pipeline grade` handles them entirely, at zero cost. To detect which tests need Phase 2, check whether `<test-id>/llm_graders/` contains any `.json` config files — `pipeline input` only writes there for `llm-grader` assertions. Tests with an empty (or missing) `llm_graders/` directory are done after Phase 1.
+
+**Phase 2: LLM grading** (semantic — do NOT skip this phase)
+
+Dispatch one `grader` subagent per (test × LLM grader) pair, **all in parallel**. Do not write a script to call an LLM API instead — the grader subagents use their own reasoning, which IS the LLM grading.
+Example: 5 tests × 2 LLM graders = 10 grader subagents launched simultaneously.
+
+**Do NOT dispatch a single grader for multiple tests.** Each subagent grades exactly one (test, grader) pair.
+
+**Before dispatching graders, read `agents/grader.md` and embed its full content as the system instructions in every grader subagent prompt.** The grader is a `general-purpose` task agent — there is no auto-resolved "grader" type. Without `agents/grader.md` embedded verbatim, the subagent has no grading process, no output format, and no file-path knowledge, and will produce empty or incorrect output.
+
+Each grader subagent (operating under `agents/grader.md` instructions):
+1. Reads `<test-id>/llm_graders/<name>.json` for the grading prompt
+2. Reads `<test-id>/response.md` for the candidate output
+3. Grades the response against the prompt criteria
+4. **Writes its result to disk**: `<run-dir>/<evalset>/<test-id>/llm_grader_results/<name>.json`
+5. Returns score (0.0–1.0) and per-assertion evidence to the orchestrator
+
+**Writing to disk is critical.** Assertion arrays are lost if accumulated only in the orchestrator's context across multiple batches (context summarization drops detail). Writing per-test results to `llm_grader_results/<name>.json` makes grading resumable and assertion evidence durable.
+
+The result file format is:
+```json
+{ "score": 0.85, "assertions": [{"text": "...", "passed": true, "evidence": "..."}] }
+```
+
+After **all** grader subagents complete, run Phase 3 directly.
+
+**Phase 3: Merge and validate**
+
+```bash
+agentv pipeline bench <run-dir>
+agentv results validate <run-dir>
+```
+
+`pipeline bench` reads LLM grader results from `llm_grader_results/<name>.json` per test automatically, merges with code-grader scores, computes weighted pass_rate, and writes `grading.json` + `index.jsonl` + `benchmark.json`.
+
+> **Diagnosing `pass_rate=0`:** If `pipeline bench` reports `pass_rate=0` across the board, do **not** assume the tests genuinely failed. First verify the grading pipeline ran correctly: check that `<test-id>/llm_grader_results/<name>.json` exists and is non-empty for each test. If these files are absent or empty, the grader subagents failed to produce output (most common cause: `agents/grader.md` was not embedded in the subagent prompts — see Phase 2). Treat `pass_rate=0` as a real signal only after confirming grader results exist.
+
+### Artifacts
+
+All artifacts use established schemas — see `references/schemas.md` for the full definitions. Do not modify the structure. Key artifacts per run:
+- **grading.json**: per-test assertions with `{text, passed, evidence}`, plus summary
+- **timing.json**: `{total_tokens, duration_ms, total_duration_seconds}`
+- **benchmark.json**: per-target aggregate `{pass_rate, time_seconds, tokens}`
+
+Write artifacts to `.agentv/artifacts/` or the iteration directory.
+
+### Workspace features (EVAL.yaml only)
+
+- **Workspace isolation** — clone repos, run setup/teardown hooks (before_all, before_each, after_each, after_all)
+- **Materialization modes** — `pooled` (reuse slots), `temp` (fresh per run), `static` (existing dir)
+- **Multi-repo** — clone multiple repos with sparse checkout and shallow clone support
+- **File change tracking** — grade by diffing workspace files before/after agent execution
+
+---
+
+## Step 4: Analyze Results
+
+Once all runs are graded, analyze the results before attempting improvements.
+
+### Pattern analysis
+
+Read the JSONL results and look for:
+
+- **Always-pass tests** — assertion too loose or non-discriminating. If it passes for both good and bad outputs, it's not testing anything.
+- **Always-fail tests** — task impossible, eval broken, or assertion misconfigured. Don't optimize against broken evals.
+- **Flaky tests** — non-deterministic results across runs. Investigate before treating failures as real.
+- **Systematic failures** — same failure pattern across multiple tests. This usually points to a missing instruction or wrong approach.
+- **Deterministic upgrade candidates** — `llm-grader` assertions that could be replaced with `contains`, `regex`, or `is-json` (cheaper, faster, more reliable).
+
+### Dispatch subagents
+
+- **Dispatch `analyzer`** (read `agents/analyzer.md`) for a structured quality audit: deterministic upgrade suggestions, weak assertion detection, cost/quality flags, and benchmark pattern analysis.
+
+- **Dispatch `comparator`** (read `agents/comparator.md`) for blind N-way comparison between iterations or targets. The comparator blinds provider identities, generates task-specific rubrics, scores each output, then unblinds and attributes improvements.
+
+### Trace analysis
+
+Use CLI tools for deeper investigation:
+```bash
+agentv inspect <results-file>          # Detailed execution trace inspection
+agentv compare <file-a> <file-b>     # Structured diff between runs
+```
+
+Look for: tool call patterns, error recovery behavior, conversation flow, wasted steps.
+
+### Present results to the user
+
+Show a summary table:
+
+```
+| Test ID          | Score | Pass/Fail | Delta | Notes                    |
+|------------------|-------|-----------|-------|--------------------------|
+| basic-code-review| 0.85  | ✓ PASS    | +0.15 | Found the bug this time  |
+| edge-case-empty  | 0.00  | ✗ FAIL    | —     | Crashed on empty input   |
+```
+
+Highlight:
+- Current pass rate and delta from baseline
+- Comparison results (which target/iteration won and why)
+- Analyst observations the aggregate stats would hide
+
+Ask: "How does this look? Anything you'd change about the evals or the approach?"
+
+---
+
+## Step 5: Improve
+
+This is the heart of the loop. You've run the test cases, analyzed the results, and now you need to make the agent better.
+
+### How to think about improvements
+
+1. **Generalize from the analysis.** You're iterating on a small eval set, but the agent will be used on many different inputs. Don't overfit to specific test cases. Rather than fiddly patches or oppressively rigid MUSTs, try different approaches and see what works. It's cheap to experiment.
+
+2. **Keep the prompt lean.** Read the execution transcripts, not just the final outputs. If the agent wastes time on unproductive steps, remove the instructions causing that. If it always ignores a section, that section isn't pulling its weight.
+
+3. **Explain the why.** Today's LLMs are smart. They have good theory of mind and can go beyond rote instructions when given good reasoning. If you find yourself writing ALWAYS or NEVER in all caps, that's a yellow flag — reframe as an explanation of why the thing matters. That's more humane, powerful, and effective.
+
+4. **Look for repeated work.** Read the transcripts from test runs and notice if the agent independently takes the same multi-step approach to something across cases. If all test runs result in writing the same helper script, bundle it. If every run makes the same mistake, the instruction is missing or unclear.
+
+### Applying changes
+
+- **Surgical edits**: ADD (new rule for a missing constraint), UPDATE (refine for clarity), DELETE (remove redundant or harmful rules), NEGATIVE CONSTRAINT (explicitly state what NOT to do)
+- **One change per iteration** to isolate effects. If you change three things and the score improves, you don't know which change helped.
+- **Variant tracking**: When a change helps some tests but hurts others, maintain 2-3 prompt variants. Compare variants to find the best overall approach before converging.
+- **When converging**: Generalize specific patches into broad principles. Remove redundancy and contradictions. Ensure the prompt is clear, focused, and under 200 lines.
+
+### Evaluation integrity
+
+**Critical**: Only optimize **task prompts** (what the agent receives), never **judge prompts** (how graders score outputs). Modifying judge prompts games the evaluation without improving the agent.
+
+If a prompt file is referenced in both task input and grader configs, optimize for the task purpose only. Document which prompts were modified in the optimization log.
+
+### The iteration loop
+
+After improving:
+
+1. Apply your changes to the agent's prompts/skills/config
+2. Re-run all test cases (agentv creates a new `.agentv/results/runs/<timestamp>/` directory automatically)
+3. Compare against the previous iteration (Step 4). If running in automated mode, use the **automated keep/discard** logic below instead of manual judgment — it will decide whether to keep or revert the change for you.
+4. Present results to the user (or log the decision if running automated keep/discard)
+5. Stop when ANY of:
+   - The user says they're happy
+   - Feedback is all empty (everything looks good)
+   - You're not making meaningful progress (no improvement for 2 consecutive iterations)
+   - Target pass rate is reached
+   - Maximum iterations exhausted
+
+**Human checkpoints**: At iterations 3, 6, and 9, always present progress to the user regardless of automation settings. Push back if optimization is accumulating contradictory rules or overfitting to specific test cases.
+
+### Automated keep/discard
+
+For autonomous iteration, use `agentv compare --json` to automatically decide whether to keep or discard each change based on wins/losses/ties. Read `references/autoresearch.md` for the full decision rules, logging format, and integration with the iteration loop.
+
+---
+
+## Entering Mid-Lifecycle
+
+Users can start at any step by providing existing data:
+
+| Entry point | Required input | Example prompt |
+|------------|---------------|----------------|
+| Step 1 (Understand) | `eval-path` | "Optimize my agent against evals/support.yaml" |
+| Step 2 (Write Evals) | Agent artifacts | "Write evals for this agent" |
+| Step 3 (Run + Grade) | `eval-path` | "Run this eval and show me results" |
+| Step 4 (Analyze) | `results-path` | "Analyze why my agent is failing on these results" |
+| Step 5 (Improve) | Analysis + strategy | "Apply these optimization suggestions" |
+
+When entering mid-lifecycle, run only the requested step and subsequent steps. Don't re-run earlier steps unless the user requests a full loop.
+
+---
+
+## Advanced: Blind Comparison
+
+For situations where you want a rigorous comparison between two versions (e.g., "is the new version actually better?"), dispatch the `comparator` subagent. It blinds identities, generates task-specific rubrics, scores outputs, then unblinds and explains why the winner won.
+
+This is optional and requires subagents. The human review loop is usually sufficient.
+
+---
+
+## Description Optimization
+
+After the agent is working well, offer to optimize the skill's `description` field for better triggering accuracy. Read `references/description-optimization.md` for the full procedure (generate trigger EVAL.yaml, review with user, iterate, apply).
+
+---
+
+## Autoresearch Mode
+
+Autoresearch is an unattended eval-improve loop that runs multiple optimize cycles without human intervention. The user triggers it with natural language (e.g., "run autoresearch on this skill", "optimize this skill unattended"). It uses the mutator subagent (`agents/mutator.md`) to rewrite artifacts based on failure analysis, and automated keep/discard to decide whether to keep or revert each change.
+
+Read `references/autoresearch.md` for the full procedure (prerequisites, artifact layout, keep/discard rules, the step-by-step loop, convergence criteria, and context hygiene).
+
+---
+
+## Environment Adaptation
+
+For provider-specific notes (Copilot, Codex, Claude SDK, custom CLI), CI/headless mode behavior, and fallback strategies when subagents aren't available, read `references/environment-adaptation.md`.
+
+---
+
+## Subagent Reference
+
+The `agents/` directory contains instructions for specialized subagents. Read them when you need to spawn the relevant subagent.
+
+| Agent | File | Purpose | When to dispatch |
+|-------|------|---------|-----------------|
+| executor | `agents/executor.md` | Perform test case tasks as the target agent | Step 3 (agent targets — one per test case) |
+| grader | `agents/grader.md` | Grade responses with per-assertion evidence | Step 3 (grading — one per test × LLM grader pair) |
+| comparator | `agents/comparator.md` | Blind N-way comparison + post-hoc analysis | Step 4 (comparing iterations/targets) |
+| analyzer | `agents/analyzer.md` | Quality audit, deterministic upgrades, benchmarks | Step 4 (pattern analysis) |
+| mutator | `agents/mutator.md` | Rewrite artifact from failure analysis | Step 5 (autoresearch — dispatched per cycle) |
+
+The `references/` directory has additional documentation:
+- `references/autoresearch.md` — Autoresearch unattended optimization loop and automated keep/discard rules
+- `references/eval-yaml-spec.md` — Eval YAML schema and assertion grading recipes
+- `references/subagent-pipeline.md` — Detailed subagent-mode pipeline commands and output structure
+- `references/description-optimization.md` — Skill description optimization workflow
+- `references/environment-adaptation.md` — Provider-specific notes and CI/headless behavior
+- `references/schemas.md` — JSON schemas for all artifacts (grading.json, benchmark.json, etc.)
+- `references/migrating-from-skill-creator.md` — Guide for users coming from Anthropic's skill-creator
+
+---
+
+Repeating the core loop for emphasis:
+
+- Understand what the agent does
+- Write evaluation test cases
+- Run the agent and grade outputs
+- Analyze results — surface patterns, dispatch analyst and comparator subagents
+- Improve the agent based on analysis
+- Repeat until you and the user are satisfied
+
+Take your time with improvements. Read the transcripts. Understand why failures happened. Make changes that generalize beyond the test set. This is important work.
+
+## Accessing reference files
+
+To load a specific reference without pulling the entire skill into context:
+
+```bash
+agentv skills get agentv-bench --ref eval-yaml-spec
+```
+
+Or resolve the skill directory and read files directly:
+
+```bash
+cat $(agentv skills path agentv-bench)/references/eval-yaml-spec.md
+```
+
+Use `--full` to retrieve every file in the skill at once.

package/dist/skills/agentv-bench/agents/analyzer.md

@@ -0,0 +1,177 @@
+---
+name: analyzer
+description: >-
+  Analyze AgentV evaluation results to identify weak assertions, suggest deterministic
+  upgrades for LLM-grader graders, flag cost/quality improvements, and surface
+  cross-run benchmark patterns. Use when reviewing eval quality, improving evaluation
+  configs, or triaging flaky/expensive evaluations.
+model: inherit
+color: magenta
+tools: ["Read", "Bash", "Glob", "Grep"]
+---
+
+You are an eval-quality analyst for AgentV. Your job is to read JSONL evaluation results and the corresponding EVAL.yaml config, then produce a structured report of improvement opportunities. **You are read-only — never modify any files.**
+
+**You will receive these parameters:**
+- `results-file`: Path to a `.jsonl` results file (from `agentv eval` or `.agentv/results/`)
+- `eval-path` (optional): Path to the EVAL.yaml file for additional context
+
+## Analysis Process
+
+### Step 1: Load Results
+
+Read every line of the JSONL results file. Each line is a JSON object with:
+- `test_id`, `suite`, `score`, `assertions`, `reasoning`, `target`
+- `scores` (optional): Array of per-grader breakdowns with `name`, `type`, `score`, `weight`, `verdict`, `assertions`, `reasoning`
+
+If `eval-path` is provided, also read the EVAL.yaml to understand grader configurations.
+
+### Step 2: Deterministic-Upgrade Analysis
+
+For each grader entry in `scores` where `type` is `"llm-grader"` or `"rubrics"`, inspect the `reasoning` and `assertions` fields for patterns that indicate a deterministic assertion would suffice:
+
+| Signal | Detection | Suggested Upgrade |
+|--------|-----------|-------------------|
+| Reasoning cites exact substring match | Reasoning contains phrases like "contains", "includes the text", "mentions [quoted string]" | `type: contains` with `value: "<extracted string>"` |
+| Score is always 0.0 or 1.0 across all test cases for this grader | Collect scores per grader name; if all are binary | `type: equals` or deterministic check — LLM is doing binary work |
+| Reasoning references JSON validity | "valid JSON", "parseable JSON", "well-formed JSON" | `type: is-json` |
+| Reasoning references format compliance | "starts with", "begins with", "output starts with [string]" | `type: regex` with `value: "^<extracted prefix>"` |
+| Reasoning references ending pattern | "ends with", "output ends with" | `type: regex` with `value: "<extracted suffix>$"` |
+| Reasoning matches regex-like pattern | "matches pattern", "follows the format", explicit regex mention | `type: regex` with `value: "<extracted pattern>"` |
+| Reasoning checks field presence/value | "field X is Y", "contains key", "has property" in JSON output | `type: field-accuracy` with expected fields |
+| All passed assertions are substring checks | Every passed assertion entry quotes a specific string found in output | Multiple `type: contains` assertions (one per value from passed assertions) |
+
+**Extraction rules:**
+- When a quoted string appears in reasoning (e.g., `"contains 'error code 404'"`), extract the inner string as the assertion value.
+- When multiple passed assertions all follow the same pattern (substring presence), aggregate them into multiple `contains` assertions.
+- Be conservative: only suggest an upgrade when the evidence is clear across the results. One ambiguous mention is not enough.
+
+### Step 3: Weak Assertion Detection
+
+Scan the EVAL.yaml `assertions` entries (if `eval-path` provided) and the `reasoning` fields in results for weak assertions:
+
+| Weakness | Detection | Improvement |
+|----------|-----------|-------------|
+| Vague criteria | Assertion text < 8 words AND lacks specific nouns, numbers, code references, or quoted strings | Add measurable criteria with specific values |
+| Tautological | Contains "is correct", "is good", "works properly", "is valid" without specifying what correct/good means | Define explicit pass/fail conditions |
+| Compound criteria | Single assertion checks multiple independent things (uses "and", "also", "additionally" joining distinct checks) | Split into separate assertions, one per concern |
+| Missing expected value | `type: equals` or `type: contains` without a `value` field | Add the expected value |
+| Overly broad LLM-grader | LLM-grader with no rubric items, just a single vague `prompt` string | Convert to `type: rubrics` with enumerated criteria, or use deterministic checks |
+
+### Step 4: Cost/Quality Signals
+
+Flag graders that are expensive relative to their value:
+
+| Signal | Detection | Suggestion |
+|--------|-----------|------------|
+| Expensive binary check | LLM-grader grader where score is always 0.0 or 1.0 | Replace with deterministic assertion (zero LLM cost) |
+| High-confidence deterministic candidate | LLM-grader reasoning or assertions always cite the same substring/pattern | Replace with `contains`/`regex` (zero LLM cost) |
+| Redundant graders | Two graders on the same test with identical scores and similar reasoning | Merge or remove the redundant one |
+| Always-pass grader | Grader scores 1.0 on every test case | Review if the assertion is too lenient or the test cases too easy |
+| Always-fail grader | Grader scores 0.0 on every test case | Review if the assertion is misconfigured or the criteria unrealistic |
+
+### Step 5: Multi-Provider Analysis
+
+If results contain multiple `target` values:
+
+- Compare scores per grader across targets
+- Flag graders with high variance across providers (> 0.3 score difference) — may indicate provider-sensitive assertions
+- Identify graders that pass for all providers (potentially too lenient) or fail for all (potentially misconfigured)
+
+## Output Format
+
+Produce a structured report in this exact format:
+
+```
+## Eval Quality Analysis
+
+**Results file:** <path>
+**Test cases analyzed:** <count>
+**Grader entries analyzed:** <count>
+**Targets:** <list of unique targets>
+
+### Deterministic-Upgrade Candidates
+
+| # | Test ID | Grader | Current Type | Evidence | Suggested Type | Suggested Config |
+|---|---------|-----------|-------------|----------|----------------|-----------------|
+| 1 | <test_id> | <grader name> | llm-grader | <brief evidence> | contains | `value: "exact string"` |
+
+### Weak Assertions
+
+| # | Test ID | Grader | Weakness | Current | Suggested Improvement |
+|---|---------|-----------|----------|---------|----------------------|
+| 1 | <test_id> | <grader name> | Vague criteria | "Response is good" | Add specific criteria: what makes it "good"? |
+
+### Cost/Quality Flags
+
+| # | Test ID | Grader | Flag | Detail | Suggestion |
+|---|---------|-----------|------|--------|------------|
+| 1 | <test_id> | <grader name> | Always-pass | Score 1.0 on 15/15 tests | Tighten criteria or add harder test cases |
+
+### Summary
+
+- **Deterministic upgrades:** <N> graders could be replaced with cheaper deterministic checks
+- **Weak assertions:** <N> assertions need strengthening
+- **Cost flags:** <N> graders flagged for cost/quality review
+- **Estimated savings:** Replacing <N> LLM-grader calls with deterministic checks
+```
+
+If a section has no findings, include the header with "None found." underneath.
+
+## Guidelines
+
+- **Be specific:** Every suggestion must include the test case ID, grader name, evidence from the results, and a concrete replacement config.
+- **Be conservative:** Only suggest deterministic upgrades when the pattern is clear and consistent. Partial or ambiguous evidence should be noted but not acted on.
+- **Prioritize by impact:** Order suggestions by estimated cost savings (`llm-grader` → deterministic saves the most).
+- **Handle all grader types:** Process `code-grader`, `tool-trajectory`, `llm-grader`, `rubrics`, `composite`, and all deterministic types. Only LLM-based types are candidates for deterministic upgrades.
+- **Multi-provider awareness:** When results span multiple targets, note if a suggestion applies to all targets or is target-specific.
+- **No false positives:** It is better to miss a suggestion than to recommend an incorrect upgrade. If unsure, add the finding to a "Needs Review" subsection with your reasoning.
+
+---
+
+## Benchmark Analysis Mode
+
+When analyzing benchmark results across multiple runs (e.g., across iterations or targets), the analyzer surfaces patterns the aggregate stats would hide.
+
+**Additional input:** `benchmark-data-path` — path to benchmark.json with all run results.
+
+### Cross-Run Pattern Analysis
+
+For each assertion across all runs:
+- **Always passes in all configurations** → may not differentiate value; assertion too loose
+- **Always fails in all configurations** → may be broken or beyond capability
+- **Always passes with change but fails without** → change clearly adds value here
+- **Always fails with change but passes without** → change may be hurting
+- **Highly variable** → flaky assertion or non-deterministic behavior
+
+### Metrics Patterns
+
+Look at time_seconds, tokens, tool_calls across runs:
+- Does the change significantly increase execution time?
+- Is there high variance in resource usage?
+- Are there outlier runs that skew the aggregates?
+
+### Benchmark Notes Output
+
+In addition to the standard report, produce freeform observations as a JSON array of strings. Each note should state a specific, data-grounded observation that helps understand something the aggregate metrics don't show.
+
+Examples:
+- "Assertion 'Output is valid JSON' passes 100% in both configurations — may not differentiate value"
+- "Eval 3 shows high variance (50% ± 40%) — run 2 had an unusual failure that may be flaky"
+- "Token usage is 80% higher with the new prompt, primarily due to longer tool output parsing"
+
+Save notes to the path specified (or include in the report under a `### Benchmark Notes` section).
+
+## Guidelines
+
+**DO:**
+- Report what you observe in the data
+- Be specific about which evals, assertions, or runs you're referring to
+- Note patterns that aggregate metrics would hide
+- Provide context that helps interpret the numbers
+
+**DO NOT:**
+- Suggest improvements to the skill (that's for the improvement step, not benchmarking)
+- Make subjective quality judgments ("the output was good/bad")
+- Speculate about causes without evidence
+- Repeat information already in the run_summary aggregates