Tier-2 Review: issue-title (Penpot OpenCode Skills)

What We Attempted

We attempted to validate the issue-title skill as a self-contained, executable unit. The skill claims to derive a clear, well-formatted GitHub issue title from a description body, with specific mood rules (present-tense for bugs, imperative for features) and mandatory inclusion of the "where" (UI/module location). We ran the standard test harness: install verification, smoke invocation, and output-shape checks.

What Failed

0 tests passed, 1 partial, 1 skipped.

• Install (skipped): No install command is documented in SKILL.md. There is no package manager reference, no dependency list, no setup script, and no mention of how this skill is meant to be loaded into an agent runtime. The skill is a prompt/instruction file only — there is nothing to install.
• Smoke invocation (partial): The skill describes a text transformation task, but provides no CLI, API, or executable entry point. We could not invoke it programmatically. A minimal invocation would require an external LLM or a script that reads the description and applies the rules — neither is defined within the skill. We could only partially validate that the rules are internally coherent (e.g., mood and location requirements are stated), but we could not observe any output.

What We Observed

The SKILL.md content is short and focused: it specifies output format rules (mood, tense, "where" inclusion) but lacks any operational scaffolding. There is no example input/output pair, no edge-case handling (e.g., empty descriptions, extremely long text, ambiguous modules), and no fallback behavior. The skill reads as a high-level style guide rather than a runnable unit.

Critically, the skill does not state how it integrates with a host system. Is it meant to be a system prompt for an LLM? A template for a human? A function that takes a string and returns a string? The absence of an interface means the skill cannot be tested, versioned, or reused across environments without significant external interpretation.

Rating Acknowledgement

The composite score of 4.2 / 5.0 — with strong marks on trigger clarity (4.5), output specificity (4.5), scope precision (4.5), moderate self-containment (3.5), and solid reusability (4.0) — reflects the design quality of the instruction content. However, this rating is theoretical until a physical re-run resolves the failures above. The skill's high scores assume a runtime that can interpret natural-language instructions and apply them to text. That runtime is not provided, documented, or referenced.

Until the skill is packaged with an explicit invocation method (e.g., a CLI wrapper, an OpenCode plugin entry, or a documented prompt template with a test harness), the "self-containment" dimension should be considered over-scored. The skill is not self-contained in any operational sense.

Is the Skill Still Valuable in Principle?

Yes — in principle, the underlying idea is sound and useful. Deriving consistent GitHub issue titles is a real pain point, and the rules (present-tense for bugs, imperative for features, mandatory "where") are concrete and actionable. If paired with a thin executable layer — say, a Python script that calls an LLM with this prompt, or an OpenCode action that accepts a description and returns a title — this skill would likely earn its 4.2 score in practice.

But as it stands, the skill is a specification, not a skill in the executable sense. Its value is contingent on an external interpreter. That is not a fatal flaw — many prompt-based skills work this way — but it must be acknowledged. The rating should be read as: "Well-designed instructions, awaiting a runtime wrapper to become operational." Until that wrapper exists, treat the 4.2 as an upper bound, not a verified result.

Bottom line: Honest signal — the skill has good bones, but it is not yet testable. Re-run after adding an entry point and example I/O.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-09-05.

Overall: broken. 0 tests passed, 1 partial, 1 skipped; key blocker: the skill is a prompt-only definition with no executable interface or documented install procedure.

Test	Status	Notes
install	skipped	No install command is documented in SKILL.md; the skill is a prompt/instruction file with no package manager or dependency installation steps.
smoke-invocation	partial	SKILL.md describes a text transformation task but provides no CLI, API, or executable entry point; a minimal invocation would require an external LLM or script, which is not defined in the skill.