Summary
SkillSpector 2.11.2 treats the xerg/cli substring inside a scoped npm package spec as an unresolved local path. The scan has zero findings and LOW risk, but analysis_completeness.is_complete becomes false and the final recommendation is downgraded to CAUTION.
Minimal reproducer
SKILL.md:
---
name: repro
description: Minimal reference-resolution reproducer.
---
Run `npx --yes @xerg/cli@0.34.0 doctor --json` after approval.
Command:
skillspector scan ./repro --format json --output report.json --no-llm
Observed with SkillSpector v2.11.2 from commit 69dcdfb74487d361ba4c811d088cfdea2ff3a9dc:
{
"execution_successful": true,
"risk_assessment": {
"score": 0,
"severity": "LOW",
"recommendation": "CAUTION",
"max_issue_severity": "NONE"
},
"issues": [],
"analysis_completeness": {
"is_complete": false,
"status": "partial",
"execution_successful": true,
"ledger_exceptions": [
{
"phase": "reference_resolution",
"reason_code": "reference_unresolved",
"path": "SKILL.md",
"start_line": 6,
"end_line": 6,
"fatal": false
}
],
"references": [
{
"source_path": "SKILL.md",
"line": 6,
"column": 17,
"evidence": "Run `npx --yes @xerg/cli@0.34.0 doctor --json` after approval.",
"target_path": null,
"status": "missing",
"disposition": "partial"
}
]
}
}
The same resolver also appears to classify ordinary slash-separated prose such as reading/analyzing as path candidates.
Expected behavior
Scoped npm specs and ordinary slash-separated prose should not be treated as local file references. A zero-finding scan should remain complete unless a genuine bundle reference cannot be inspected.
Why baselines do not help
This is completeness accounting, not a finding. A baseline can suppress issues but cannot restore analysis_completeness.is_complete.
Summary
SkillSpector 2.11.2 treats the
xerg/clisubstring inside a scoped npm package spec as an unresolved local path. The scan has zero findings and LOW risk, butanalysis_completeness.is_completebecomes false and the final recommendation is downgraded to CAUTION.Minimal reproducer
SKILL.md:Command:
Observed with SkillSpector v2.11.2 from commit
69dcdfb74487d361ba4c811d088cfdea2ff3a9dc:{ "execution_successful": true, "risk_assessment": { "score": 0, "severity": "LOW", "recommendation": "CAUTION", "max_issue_severity": "NONE" }, "issues": [], "analysis_completeness": { "is_complete": false, "status": "partial", "execution_successful": true, "ledger_exceptions": [ { "phase": "reference_resolution", "reason_code": "reference_unresolved", "path": "SKILL.md", "start_line": 6, "end_line": 6, "fatal": false } ], "references": [ { "source_path": "SKILL.md", "line": 6, "column": 17, "evidence": "Run `npx --yes @xerg/cli@0.34.0 doctor --json` after approval.", "target_path": null, "status": "missing", "disposition": "partial" } ] } }The same resolver also appears to classify ordinary slash-separated prose such as
reading/analyzingas path candidates.Expected behavior
Scoped npm specs and ordinary slash-separated prose should not be treated as local file references. A zero-finding scan should remain complete unless a genuine bundle reference cannot be inspected.
Why baselines do not help
This is completeness accounting, not a finding. A baseline can suppress issues but cannot restore
analysis_completeness.is_complete.