[rig-tasks] Add 10 rig samples — 2026-10-09

[github-actions]

Summary

Added 10 new rig sample files to skills/rig/samples/.

#	File	Description	Typecheck
1	531-commit-type-classifier.md	Classify recent commit subjects into conventional-commit types via p.bash + s.enum/s.record	pass
2	532-tsconfig-deep-validator.md	Validate tsconfig.json compilerOptions via p.read against a deeply nested s.object schema	pass
3	533-repair-addon-strict-output.md	Strict s.object output with repair() addon and higher maxTurns	pass
4	534-healthcheck-timeout-status.md	p.bash health-check with timeout() addon returning status/durationMs	pass
5	535-lint-test-gate-workflow.md	workflow() chaining lint-check, test-check, and a gate agent sequentially	pass
6	536-glob-tagger-fanout.md	p.glob discovery fanning out to a tagger subagent, aggregated into s.record(s.array(path))	pass
7	537-line-count-report-writer.md	p.bash line counting + p.writeOutput markdown report side effect	pass
8	538-semver-satisfies-tool.md	Pure-logic defineTool for semver range checks + p.read of package.json	pass
9	539-todo-fixme-parallel-workflow.md	workflow() using parallel() to fan out TODO/FIXME counting agents, then aggregate	pass
10	540-steering-nullable-commit.md	[steering(), repair()] addon chain with s.nullable output field	pass

All 10 programs passed node skills/rig/rig.ts <file> --typecheck.

Typecheck failures

None — all 10 tasks passed typecheck on the first attempt.

Tasks run

• (reused) Run p.bash to fetch recent commits, classify each commit message into an s.enum of conventional-commit types, and return an s.record keyed by type mapping to matching commit subjects.
• (reused) Use p.read to load tsconfig.json and validate it against a deeply nested s.object schema (compilerOptions with optional strict/target/paths as s.record(s.string, s.array(s.string))), outputting a validation report.
• (reused) Build an agent with a strict s.object output schema and attach the repair() addon with a higher maxTurns to demonstrate automatic re-prompting on malformed output.
• (reused) Run a p.bash health-check command with an explicit timeout() addon, returning s.object with status: s.enum('ok','timeout','error') and durationMs: s.number.
• (reused) Compose a workflow (not a single agent) chaining three agents sequentially: lint-check, test-check, and a gate agent combining both into an approved/blocked decision.
• (reused) Use p.glob to find source files, fan out to a tagger subagent that assigns topical tags from an s.enum set, then aggregate into an s.record(s.string, s.array(s.string)) of tag to file paths.
• (new) Use p.bash to count lines per file for all *.ts files under src/, then p.write to persist a generated markdown report; output s.object({ reportPath, totalLines, fileCount }).
• (new) Define a custom defineTool (checkSemverSatisfies) computing version-range satisfaction via pure logic; use p.read on package.json and output an array of per-dependency results.
• (new) Compose a workflow using parallel() to run TODO/FIXME counting agents concurrently, then aggregate into totals.
• (new) Build an agent with [steering(), repair()] addons and an s.nullable output field, reading the latest commit via p.bash.

Notes

The task-generator and rig-expander custom subagents failed on every invocation with 400 Cannot translate Copilot request feature 'frequency_penalty' between Responses and Chat Completions — an infra-level model routing incompatibility unrelated to this task. Task generation and program authoring were performed directly against skills/rig/SKILL.md and its focused references instead, so today's run has reduced subagent-delegation coverage compared to typical runs.

Generated by Daily Rig Task Generator · copilot · auto · 179.9 AIC · ⌖ 17.9 AIC · ⊞ 9.4K · ◷

[github-actions]

🧠 Matt Pocock Skills Reviewer is reviewing this pull request using Matt Pocock's engineering skills...