Ts Code Validation Gate

TS Code Validation Gate

Use this skill after implementation appears complete and before code review.

This skill enforces a single-pass evidence capture workflow across quality, test, and security checks, then blocks review if unresolved findings remain.

When to use

Use this skill when prompts imply:

• “final check before review”
• “run validation gate”
• “collect lint/test/security evidence before PR review”
• “fail closed until remediation is complete”
• “do one pass and tell me what still blocks review”

What this gate does

1. Ensures .agents/.code-validation-gate/ is ignored via .git/info/exclude.
2. Creates a timestamped run folder:

• .agents/.code-validation-gate/YYYY-MM-DD_HHMMSS/

1. Runs checks in one pass and writes evidence files:

• Formatting/linting auto-detection:
• Biome when biome.json or biome.jsonc exists at project root
• ESLint when root eslint config exists
• Prettier when root prettier config exists
• Unit tests (bail first, failures-focused)
• Semgrep auto scan
• osv-scanner recursive scan
• fallow compact audit with fail-on-issues

1. Emits machine-readable status JSON.
2. If any check fails, creates remediation-ledger.md and needs-remediation.flag.
3. On the next run, fails closed if the most recent unresolved ledger still contains unchecked items (- [ ]).

Required execution command

bash /mnt/skills/user/ts-code-validation-gate/scripts/run-gate.sh

Optional repository root override:

bash /mnt/skills/user/ts-code-validation-gate/scripts/run-gate.sh /path/to/repo

Evidence artifacts

The run folder contains:

• status.json
• results.tsv
• biome.txt (if biome detected)
• eslint.json (if eslint detected)
• prettier.txt (if prettier detected)
• tests.txt
• semgrep.json
• osv.md
• fallow.txt
• remediation-ledger.md (only when failures exist)
• needs-remediation.flag (only when failures exist)

Remediation policy (fail closed)

• Any failing check is blocking for review readiness.
• The agent must remediate findings and update checklist items in remediation-ledger.md to - [x].
• If a prior run remains unresolved, future runs fail closed and a new review is invalid.

Agent output contract

When reporting gate results, always provide:

• overall gate state: pass, fail, or fail-closed
• run directory path
• failed checks list
• exact evidence file paths
• remediation ledger path when present
• explicit statement about whether review is currently valid

Maintainer notes

Skill measurement guidance (including eval workflows) is documented in README.md in this skill directory, not in runtime agent instructions.

Reference

Read references/gate-contract.md when you need exact behavior and failure semantics.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: mwillbanks
• Source: mwillbanks/agent-skills
• License: Apache-2.0

Install and usage instructions live in the source repository linked above.