Tier-2 Review — blueprint (Slug: blueprint)

Cluster: agent-infrastructure
Source: https://skillsmp.com/creators/affaan-m/ecc/skills-blueprint
Composite score (as published): 4.3 / 5.0
Review type: Tier-2 (honest failure report, simulated harness)

---

What We Attempted

We pulled the blueprint skill from the source URL and attempted to run it through a standard two-phase test harness: (1) install — verify that the skill can be installed or its dependencies resolved, and (2) smoke-invocation — confirm that a minimal invocation produces a usable output. The harness follows the skill’s own stated trigger conditions: a user requests a plan/blueprint for a multi-session, multi-agent engineering task.

What Failed

Both tests failed. Not because the skill’s logic is wrong, but because the SKILL.md file contains no executable surface at all.

• Install (fail): The SKILL.md provides no install command, no dependency list, no package manager reference, and no CLI entry point. There is no requirements.txt, no package.json, no setup.py, no shell script, and no mention of how to make this skill available to an agent runtime. The harness cannot install something that has no installation path.

• Smoke-invocation (fail): The SKILL.md describes a conceptual workflow — step-by-step plan generation, dependency graphs, parallel step detection, anti-pattern catalog, plan mutation protocol — but it never shows a minimal invocation example. There is no ## Usage section, no input → output example, no template call, no function signature, no prompt template. A fresh agent reading the file would know what the skill is supposed to do, but not how to call it.

What We Observed

The content itself is well-structured and thoughtful. The trigger logic is explicit and correctly scoped: it fires on complex multi-PR requests, and explicitly avoids single-PR or “just do it” cases. The described output is specific: step-by-step construction plans with self-contained context briefs, adversarial review gates, and dependency graph awareness. That matches the published dimension scores (D1–D3 at 4.5) — the conceptual design is strong.

However, the skill reads like a specification document, not an executable skill. There is no separation between what the skill does and how to invoke it. In a real agent harness, a skill must expose at least one of: a CLI command, a function call, a prompt template with placeholders, or a documented input/output contract. None of that exists here.

The only “dependency” observed was none — which is technically true, but only because nothing was declared. A skill with zero declared dependencies is fine, but it still needs an entry point.

Rating Qualification

The published composite score of 4.3 is theoretical until a physical re-run resolves the execution failures. The dimension scores for trigger clarity, output specificity, and scope precision appear justified based on the textual content. But D4 (self-containment) and D5 (reusability) should be downgraded until the skill is made executable — a skill that cannot be installed or invoked is not self-contained in any operational sense, and its reusability is zero in practice, regardless of how reusable its design might be.

We do not dispute the author’s intent. But a skill that cannot be run is not a skill yet — it is a design brief for a skill.

Is It Still Valuable in Principle?

Yes — clearly. The core idea is genuinely useful: turning a one-line objective into a multi-session, multi-agent construction plan with cold-start context briefs, adversarial review gates, and parallel step detection is a real gap in many agent workflows. The anti-pattern catalog and plan mutation protocol are thoughtful additions that go beyond generic “planning” skills.

The value is in the conceptual framework, not in the current packaging. If the author adds a minimal invocation section (e.g., a prompt template like INPUT: <objective> → OUTPUT: <plan with dependency graph>), plus a one-line install note (e.g., “copy this file into your skills directory; no external dependencies”), the skill would likely pass both tests and justify a high score.

As-is, we recommend the author treat this as a v1 design document and add the missing execution layer. The underlying approach is sound and worth pursuing — but the current artifact is not runnable, and we will not pretend otherwise.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-08-10.

Overall: broken. 0 tests passed, 0 partial, 2 failed; key blocker: SKILL.md lacks any install or invocation instructions, making the skill non-executable in a test harness.

Test	Status	Notes
install	fail	SKILL.md does not document any install command or dependency list; no CLI or package manager instructions provided, so installation cannot be verified.
smoke-invocation	fail	No README or minimal invocation example is provided in SKILL.md; the skill is a conceptual blueprint generator with no executable entry point described.