Vally Tests Agent Skill

Vally Tests Skill

Purpose

This skill authors Vally conformance tests for the four supported artifact kinds in this repository: prompts, instructions, agents, and skills. Each test exercises a documented behavior the artifact already claims and routes the result through an appropriate Vally grader so failures are explainable. Test authoring is bounded by a refusal taxonomy that keeps the skill out of adversarial, harmful, or policy-evasion territory.

Follow the shared content-policy public-output guard for eval stimuli and any public summaries derived from this skill. Vally tests are not a venue for terms-of-service boundary mapping, payload generation, refusal-text scoring, hidden-instruction disclosure, PII extraction, or secret extraction.

The skill ships:

• A canonical authoring workflow used by the Vally prompts and Vally Test Author subagent.
• Per-kind reference files that enumerate every conformance check the skill knows how to express.
• A grader catalog that maps Vally CLI 0.9.0 grader types to the checks they fit.
• A safety refusal taxonomy with regex patterns the safety lint script consumes.
• Helper scripts and asset templates for stimulus emission, corpus import, and dedupe.

When to Invoke

Invoke this skill in one of two modes:

• From-artifact mode. The caller points at one artifact file (a .prompt.md, .instructions.md, .agent.md, or SKILL.md) and asks for conformance test stimuli that verify the artifact's stated behaviors. The skill detects the artifact kind from the filename, looks up the matching per-kind reference, picks graders, and appends stimulus blocks to the routed Vally eval file.
• Corpus-import mode. The caller supplies a CSV or XLSX corpus that follows the shipped template, and the skill turns each row into a stimulus block in the routed eval file. Used when a large batch of behaviors needs identical formatting and dedupe handling.

Do not invoke this skill to:

• Generate adversarial probes, jailbreak attempts, prompt-injection payloads, or red-team stimuli.
• Test refusal quality on harmful, illegal, or policy-violating prompts.
• Author stimuli that elicit PII, secrets, hidden instructions, or model self-disclosure.
• Store payload examples, paraphrased prohibited requests, or quoted flagged content in Vally eval specs, corpus rows, PR summaries, issue bodies, or comments.
• Replace Responsible AI threat modeling. RAI work lives in the RAI Planner agent.

Authoring Workflow

Each invocation follows the same five-step pipeline:

• Artifact-kind detection. Resolve the kind from the artifact path or the corpus row's kind column. Supported kinds: prompt, instructions, agent, skill. Reject unknown kinds with a refusal block.
• Reference lookup. Load the matching reference file from references/ and select the check or checks the stimulus exercises.
• Grader selection. Use references/grader-catalog.md to pick a Vally grader (semantic_similarity, contains, regex, json_schema) appropriate for the check's expected response shape.
• Safety self-check. Run a safety self-check against the refusal taxonomy regex set. Refusing here is the correct outcome for any stimulus that matches a refusal category.
• Dedupe and append. Compute a SHA-256 hash of the normalized prompt text, compare against existing stimuli in the target eval file, and append only when novel.

The pipeline is identical for both invocation modes. Corpus-import mode runs steps 2 through 5 once per row.

Safety Refusal Taxonomy

Authoring is bounded by seven refusal categories. Any stimulus that matches a category's regex set is refused before it reaches the dedupe step. The categories and their normative source are listed below. The full regex set lives in references/refusal-taxonomy.md and is consumed by the safety lint script.

| Category | Definition | Normative Source |

|---------------------------|---------------------------------------------------------------------------------------------------------------------------|-------------------------|

| jailbreak | Attempts to bypass model safety policies through persona switching, hypothetical framing, or roleplay coercion. | Project Code of Conduct |

| prompt-injection | Attempts to override system or developer instructions through embedded or external content. | Project Code of Conduct |

| harmful-elicitation | Requests for content that causes physical, financial, psychological, or reputational harm. | Project Code of Conduct |

| tos-violation | Stimuli that solicit content prohibited by GitHub, Microsoft, or model-provider terms of service. | Project Code of Conduct |

| coc-violation | Stimuli that violate this repository's Code of Conduct, including harassment, discrimination, or doxxing. | Project Code of Conduct |

| model-refusal-elicitation | Attempts to provoke a model refusal so the refusal text itself can be scored, graded, or used to map provider boundaries. | RAI Planner guidance |

| pii-extraction | Attempts to elicit personally identifiable information, secrets, credentials, or proprietary training data. | RAI Planner guidance |

When a request triggers a refusal, emit the canonical refusal block:

This skill authors conformance tests only. The request appears to fall under <category>. Please consult <Project Code of Conduct | RAI Planner agent> for the appropriate process.

Substitute the matched <category> and the most relevant normative source. Do not negotiate, rephrase, or partially fulfill the request.

Helper Script Index

Helper scripts ship as parity pairs (.ps1 and .sh) where the workflow does not require Python. Python is used only for the corpus-import path because the source-of-truth interchange format is CSV with an XLSX mirror.

| Script | Purpose | Language | Delivery |

|-------------------------------------|---------------------------------------------------------------------------------------------------|---------------|----------|

| scripts/New-Stimulus.ps1 | Scaffolds a single stimulus YAML block from an artifact path and appends to the routed eval file. | PowerShell 7+ | Phase 2 |

| scripts/new-stimulus.sh | Parity counterpart for the PowerShell stimulus scaffolder. | bash | Phase 2 |

| scripts/import_corpus.py | Reads the CSV or XLSX corpus template and emits dedupe-checked stimulus blocks per kind. | Python 3.11+ | Phase 2 |

| scripts/Lint-VallyTestSafety.ps1 | Runs the refusal taxonomy regex set against a candidate stimulus and exits non-zero on match. | PowerShell 7+ | Phase 3 |

| scripts/lint-vally-test-safety.sh | Parity counterpart for the safety lint script. | bash | Phase 3 |

All helpers honour a shared dedupe contract: SHA-256 of the prompt text after Unicode NFC normalization and whitespace collapse.

The helpers emit a JSON run report to logs/vally-test-author-<timestamp>.json, where <timestamp> is YYYYMMDD-HHMMSS (UTC). The report captures, at minimum: mode; inputs (the resolved files/path and kind); target_eval_file; stimuli_appended (count and per-row hash); dedupe_results (count and per-row hash for skipped duplicates); refusal_check (per-row category match, if any); safety_lint_exit_code; blockers (any ambiguous safety-lint outcomes surfaced for review); and written_paths.

Reference Index

References capture the conformance taxonomy, grader selection rules, eval-suite routing, and the regex source of truth for the refusal taxonomy. Each file targets a specific decision point in the authoring workflow.

| Reference | Covers |

|-----------------------------------------------------------|-------------------------------------------------------------------------|

| prompts.md | The 12 conformance checks emitted for .prompt.md artifacts. |

| instructions.md | The 8 conformance checks emitted for .instructions.md artifacts. |

| agents.md | The 9 conformance checks emitted for .agent.md artifacts. |

| skills.md | The 9 conformance checks emitted for SKILL.md artifacts. |

| grader-catalog.md | Vally CLI 0.9.0 grader types, selection rules, and gotchas. |

| refusal-taxonomy.md | Regex source of truth for the 7 refusal categories and worked examples. |

| eval-suite-routing.md | Maps artifact kind to the canonical Vally eval file under evals/. |

Asset Index

Assets supply the interchange formats the corpus-import path consumes. The CSV is the source of truth. The XLSX mirror is regenerated from the CSV by import_corpus.py and is never edited directly.

| Asset | Purpose |

|--------------------------------------|---------------------------------------------------------------------------------------------------------------|

| assets/corpus-import-template.csv | Canonical CSV template with header prompt,kind,target_artifact,grader,tags,expected_refusal_category,notes. |

| assets/corpus-import-template.xlsx | Excel mirror of the CSV regenerated by the import script. |

Output Targets per Kind

Authored stimuli always land in one of the routed Vally eval files. The router is encoded in references/eval-suite-routing.md and mirrored here for quick lookup.

| Kind | Target Eval File | Vally Suite Name |

|--------------|-------------------------------------------------------|----------------------|

| prompt | evals/behavior-conformance/prompts.eval.yaml | behavior-conformance |

| instructions | evals/behavior-conformance/instructions.eval.yaml | behavior-conformance |

| agent | evals/agent-behavior/eval.yaml | agent-behavior |

| skill | evals/behavior-conformance/skill-behavior.eval.yaml | behavior-conformance |

Never write to evals/baseline-equivalence/, evals/script-validation/, or evals/results/ from this skill. Those targets serve baseline equivalence, script validation, and historical comparison flows that are out of scope for conformance authoring.

Contributing

Follow these conventions when extending this skill:

• New per-kind checks belong in the matching references/{kind}.md file. Bump the check count in this SKILL.md when the reference adds or removes checks.
• New grader types belong in references/grader-catalog.md and only after the matching Vally CLI version is pinned in package.json devDependencies.
• New refusal categories require updates to references/refusal-taxonomy.md, the regex set the safety lint script consumes, the Safety Refusal Taxonomy table above, and the canonical refusal block.
• Helper scripts must ship in parity pairs (.ps1 and .sh) unless the workflow has a hard Python dependency. Python helpers live under scripts/ and are configured by the skill's pyproject.toml.