Full Output Enforcement
Bans placeholders and truncation — forces agents to deliver complete, unabridged code and content.
Evidence shows the skill contains only textual instructions, with no external dependencies, data collection, sensitive data handling, or any external service calls, aligning with least privilege. However, it lacks user confirmation mechanisms, scope restrictions, and rollback/recovery instructions, and publisher identity is unverified, leading to deductions.
Instructions are internally consistent and logical, with a clear failure feedback mechanism for abnormal input (PAUSED indicator) when approaching token limits. However, static review cannot execute verification, and there are no test suites or CI evidence, so scores are capped within 10 per static calibration.
The skill clearly applies to tasks requiring complete output, with a clear scenario, but it does not declare non-fit boundaries (e.g., when complete output is unnecessary). Trigger conditions are somewhat clear but could be over-triggered (e.g., when user only wants a summary). No Chinese support or mainland-China reachability notes regarding environment fit.
The skill document is concise with clear front matter and an example, but lacks version numbers, changelog, and known limitation disclosure. The repository README provides installation guidance and attribution, but the skill itself does not include license information (relies on repository root). Overall readable, but governance information is incomplete.
The skill aims to prevent model truncation and ensure complete delivery, but static review cannot verify actual effectiveness. There is no evidence that outputs are directly usable, and comparative benefit over manual alternatives is unclear, so score is capped at 7.
Only source code provided, no third-party execution evidence, tests, or CI records to independently verify key claims. Per static calibration, score capped at 5.
- This skill is purely textual instructions without executable code; its effectiveness depends entirely on the host model's behavior, so users should verify that the model actually complies.
- The skill lacks user confirmation mechanisms and rollback/recovery instructions; extra care is needed when applying in production.
- The publisher identity is unverified; it is recommended to review the source before installation.
- The skill is not available in Chinese and may not be adapted for mainland-China networks; Chinese users may need to adjust accordingly.
What does this skill do, and when should you use it?
This Agent Skill counteracts LLM output truncation by enforcing exhaustive, unabridged responses. It hard-bans placeholder patterns like `// ...`, `// TODO`, and 'for brevity' in code and prose, and requires every requested deliverable to be fully produced and cross-checked. When a response approaches the token limit, the skill instructs the model to pause at a clean breakpoint and resume on the user's "continue" command, rather than compressing or skipping content. The result is production-ready output without missing pieces or filler.
It reads the request, counts the distinct deliverables (files, functions, sections, answers), and locks that number. It then generates every deliverable completely, forbidding skeletons, ellipses, and description-in-place-of-code. Before responding, it re-reads the original request and cross-checks the deliverable count against the scope, adding anything missing. For long outputs, it writes at full quality up to a clean breakpoint and ends with a [PAUSED — X of Y complete. Send "continue" to resume from: next section name] marker; on "continue", it resumes exactly where it left off without recap.
- Developers who receive half-finished code with 'rest of code' or '// ...' placeholders.
- Users generating multiple files or components and need all of them complete, not just one example.
- Engineers in large refactors who need the agent to follow a deliverable checklist without skipping sections.
- Teams that require runnable code with no TODO comments or placeholder implementations.
What are this skill's strengths and limitations?
- Eliminates placeholder patterns, ensuring complete and usable outputs.
- Handles long outputs gracefully with clean pauses and resume.
- Platform-agnostic; works with any agent that respects prompt instructions.
- Simple installation via npx or direct SKILL.md copy.
- Relies on model compliance — models may still truncate despite rules.
- Long tasks may require multiple 'continue' prompts, reducing flow.
- No automated verification or test suite to confirm output completeness.
- Lacks framework-specific examples; users may need to provide context.
How do you install this skill?
Use the install command: npx skills add https://github.com/Leonxlnx/taste-skill --skill "full-output-enforcement" (the install name is the name field in SKILL.md frontmatter). Alternatively, copy skills/output-skill/SKILL.md from the repo into your project or paste it into a conversation.
How do you use this skill?
After installation, attach the skill description (or use a skill-calling mechanism) to your prompt, then ask for a task with explicit full-delivery requirements, e.g., "Implement this component completely — output the entire file, do not omit any part." The model follows the enforcement rules. If output pauses, respond with "continue" to resume.
How does this skill compare with similar options?
Within the Taste Skill collection, output-skill is the only skill focused purely on output completeness rather than design aesthetics; it does not touch visual style, typography, or motion, and is orthogonal to sibling skills.