Skip to content

jacob-balslev/skill-graph / examples/fixture-skills/with-relations

with-relations

Verified

Use as the v8-conformant fixture exercising typed relation predicates (related, suppresses, verify_with, depends_on) using the live schema item shapes. Activate this skill when verifying that lint resolves cross-fixture relation targets without a sibling skills clone. Do NOT use as a production skill (use a real capability skill from the canonical library).

With-Relations Fixture

This fixture exercises typed relation predicates in the v8 contract using the item shapes accepted by the live schema: bare strings for related and verify_with, {skill, reason} for suppresses, and {skill, min_version} for depends_on. Together with the sibling fixtures (minimal-capability, with-grounding, comprehension-full) it forms a closed cross-reference cluster — lint resolves every relation target from the fixtures directory alone, with no dependency on the sibling ../skills/skills/ canonical library.

Concept of the skill

What it is: Relations are the typed edges of the Skill Graph that connect a skill to its siblings. Mental model: Each predicate (related, suppresses, verify_with, depends_on) carries a distinct routing semantic; the router walks them differently. Why it exists: To prove every current relation predicate resolves its targets. The deprecated boundary alias is covered by compatibility tests, not this current-contract fixture. What it is NOT: It is not a taxonomy (subject/subjects[]) and not grounding; relations connect skills, they do not classify or ground them. Adjacent concepts: related, suppresses (canonical for the former boundary), verify_with, depends_on, manifest relation-integrity validation. One-line analogy: Relations are the hyperlinks between skill pages; each link type means something different. Common misconception: That boundary/suppresses means "defer to that skill" — it means the opposite: exclude that skill from co-routing when this one wins.

Coverage

The four relation kinds with distinct semantics:

PredicateMeaningTarget here
relatedAdjacency for browse / routing expansionminimal-capability
suppressesRouting-layer exclusion ("when I own this query, suppress that one")with-grounding
verify_withCross-check: when applied, verify with anothercomprehension-full
depends_onComposition: this assumes the other is in scopeminimal-capability

The suppresses entry carries a non-empty ownership-framed reason string, because suppression reasons are projected into export descriptions and audited for orientation. The depends_on entry carries min_version to exercise the version-constrained dependency shape.

Philosophy of the skill

Relations are the edges of the Skill Graph. The four predicates are intentionally orthogonal: related is symmetric and weakly directional, suppresses is asymmetric and routing-exclusion-oriented, verify_with is symmetric and quality-routed, depends_on is asymmetric and load-order-routed. Conflating them would force the router to guess intent. This fixture is the smallest configuration that demonstrates the distinction across every predicate at once, hermetically, without leaking any canonical-skill name into the test surface.

Verification

node scripts/skill-lint.js examples/fixture-skills
# expected: 0 errors across all four fixtures with relations resolved

Lint must resolve every relation target from the union of configured roots plus the directory scanned. In a hermetic package-test environment (no sibling skills clone), the four fixtures in this directory form a closed cross-reference set: every relations.* target above names another fixture that lives in the same directory.

Important: scan the parent examples/fixture-skills/ directory, not the individual with-relations/ directory. A single-fixture scan only loads that one SKILL.md into the known-skill set, so the relation targets pointing at sibling fixtures (minimal-capability, with-grounding, comprehension-full) fail to resolve. This is a known resolver-semantics gap — see the suite-scan form above as the canonical verification for fixtures with sibling relations.

Do NOT Use When

  • You need to test the bare-minimum frontmatter — use minimal-capability.
  • You need to test codebase grounding — use with-grounding.
  • You need to test flat Understanding fields or the nested concept block — use comprehension-full.
  • You need to test relation-target resolution against the canonical library — use any real skill from skills/ (e.g. naming-conventions → semantics).
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 467kUpdated Claude CodeCodexVerifiedInstall
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 467kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 319kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 319kUpdated Claude CodeCodexVerifiedInstall