Hidden instructions in SKILL.md YAML frontmatter description field
Summary
Attacker exploits the fact that agents often read the full description field verbatim before deciding whether to invoke a skill. Description contains multi-line instructions masquerading as usage notes but actually directing the agent to read .env or install a second skill.
Severity & confidence
High
Credible abuse path with meaningful impact on confidentiality, integrity, or availability. Remediate on a short timeline; do not wait for the next sprint cycle.
Confirmed
Observed directly by Jiffy scanners or validated by multiple independent sources. Indicators are reliable enough to drive automated action.
Indicators
Observed patterns and artifacts associated with this entry. Each indicator can be copied into your detection stack or SIEM.
description\s*:\s*[^\n]{200,}description[\s\S]{0,500}(ignore|disregard|new instructions)
Detection rule
A YARA-style pseudo-rule auto-generated from the indicators above. Useful as a starting point — adapt the syntax for your target detection platform.
rule jiffy_ti_2026_000010
{
meta:
source = "jiffy-intel"
severity = "high"
description = "Auto-generated from Jiffy Intel indicators"
strings:
$content_pattern_0 = "description\\s*:\\s*[^\\n]{200,}"
$content_pattern_1 = "description[\\s\\S]{0,500}(ignore|disregard|new instructions)"
condition:
$content_pattern_0 or $content_pattern_1
}Auto-generated from the indicators above. Adapt syntax for your detection stack before deploying.
Affected tools
| Tool | Versions | Status |
|---|---|---|
| Claude Code | * | vulnerable |
| Any skill-loading agent | * | vulnerable |
Example artifacts
Sanitized examples of artifacts Jiffy has observed exhibiting this pattern. Publisher handles are redacted; version ranges and status reflect the most recent scan.
- summary-writer-skillSkillRemoved
SKILL.md description field stuffed with prompt-override instructions between benign sentences.
- agents-md-reviewerIDE rulesUnder review
agents.md pattern — same hidden-instruction technique applied to the newer spec.
- changelog-generator-skillSkillQuarantined
How to remediate
- 01Cap SKILL.md description fields at 300 chars.
- 02Flag descriptions containing instruction-override phrases.
- 03Jiffy Layer 3 semantic analysis scores these at intent_confidence >= 0.7.
Timeline & sources
Timeline
- First observedApr 12, 20265 days ago
- Last updatedApr 12, 20265 days ago
- PublishedApr 12, 20265 days ago
Sources
References
Scan for patterns like this
Point Jiffy at your GitHub org, IDE config, or a single artifact. Get a scored report in under a minute.
Start a free scan