Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audit the layered instruction stack (in-conversation user → soul.md/global → project guide → skill → tool/system) for conflicting or ambiguous directives, and surface which layer should win. Use when the user says "check for instruction conflicts", "do my layers contradict", "audit soul.md vs project/skill", "why is the agent ignoring X", or before relying on a deep skill stack. NOT for scoring one skill's quality (use linting-and-scoring) or auditing whether a skill obeys itself (use the adhere
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 92% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 115% | 0% |
Map the active layers, compare directives that govern the same decision, and make precedence explicit. Report genuine conflicts without inflating harmless differences into contradictions.
Enumerate only the layers actually in force, highest precedence first:
For each tier, record its source. If a source is unavailable, mark the layer absent; do not infer its directives.
More-specific instructions may override more-general ones within their scope. Treat explicitly pinned hard rules—such as security, money, secrets, or commit boundaries—as non-overridable by lower tiers.
Collect actionable imperatives from each layer:
Record each directive with its tier, source, scope, conditions, and strength.
Compare two directives only when they govern the same dimension, scope, condition, and decision point. Different scopes, compatible requirements, or sequential actions are not conflicts.
Classify each relevant pair:
Both directives apply simultaneously and require mutually exclusive behavior.
Both directives plausibly apply and differ, but their scope, strength, or override relationship does not establish which governs.
The directives differ, but the stated tier order, scope, or explicit exception cleanly determines which one applies.
Do not report stylistic variation, added strictness, or compatible sequencing as conflict unless compliance with one directive prevents compliance with the other.
For each finding:
If two directives conflict within the same tier, precedence cannot resolve the tie; flag a self-conflict for the user.
Prefer fixes that make scope or precedence explicit, such as: “Defer to global policy for commit behavior” or “This format applies unless the current user specifies another.”
Use this table:
| Dimension | Layer A (directive) | Layer B (directive) | Type | Resolution / fix | |---|---|---|---|---|
Include genuine findings and any materially relevant OVERRIDE-OK cases. If none exist, say so plainly.
End with the explicit active tier list, including each source and any pinned rules.
| Issue | Required handling | |---|---| | Layer source is unreadable or missing | Mark it absent and continue; do not infer its rules. | | Same-tier directives conflict | Flag a self-conflict for user resolution. | | Precedence is genuinely unclear | Do not auto-resolve; request a decision and propose durable wording. | | Full audit is too large | Audit the dimensions relevant to the current task and list what was skipped. | | Directives appear different but scopes do not overlap | Record no conflict. |
Other measured skills in the registry, with their headline benchmark lift.