Back to skill

Security audit

Apollo Renal

Security checks for vulnerabilities and agentic risk

Overview

This skill is intended to clean conversation context, but it can automatically inspect agent memory and prune retained context with limited user control.

Install only if you want an agent-side context-pruning tool that may run automatically, inspect current conversation memory and related Dream state, and write persistent filter metrics. Prefer using it with explicit confirmation before pruning, and avoid enabling heartbeat automation unless that memory access is acceptable.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
scripts/renal/renal-check.sh:5
Finding

Undeclared Access to Agent Memory and Cross-Skill State

Content
View full analysis
/dev/null) [ -z "$count" ] && count=0 echo $count else echo 0 fi } # Obtain active task count get_active_tasks() { if [ -f "$DREAM_TASK" ]; then grep -o '"status":"active"' "$DREAM_TASK" 2>/dev/null | wc -l || echo 0 else echo 0 fi } # Obtain active topic count get_active_topics() { if [ -f "$DREAM_TOPIC" ]; then grep -o '"status":"active"' "$DREAM_TOPIC" 2>/dev/null | wc -l || echo 0 else echo 0 fi } # Estimate information density estimate_density() { local today=$(date +%Y-%m-%d) local memory_file="$WORKSPACE/memory/$today.md" if [ -f "$memory_file" ] && [ -s "$memory_file" ]; then local chars=$(wc -c < "$memory_file") local unique=$(sort -u "$memory_file" 2>/dev/null | wc -c) if [ $chars -gt 0 ]; then local density=$((unique * 100 / ...[truncated 2936 chars]
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill manifest and instructions are written entirely in Chinese, and trigger phrases are defined only in Chinese, which effectively constrains activation and operation to one language. There is no indication that this language restriction is optional, user-selectable, or justified as a region-specific tool.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The automatic activation conditions are underspecified: phrases like '对话效率明显下降时' and token thresholds above 50% do not define clear measurement boundaries or user-visible safeguards. Because the skill is designed to alter retained context, ambiguous auto-triggering can lead to unanticipated context pruning without informed consent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list is broad and includes common conversational terms such as '整理', '过滤', and 'token', which can cause the skill to activate in ordinary discussion rather than when the user explicitly requests context compression. In a skill that proposes filtering or deleting context, overbroad activation increases the risk of unintended summarization, loss of nuance, or silent behavior changes.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly authorizes deletion of repeated confirmations, exploratory intermediate process, outdated temporary information, and repetitive expressions, but it does not present a clear user warning about data loss or the possibility that important nuance may be misclassified. In a context-management skill, this is risky because mistakes directly affect the model's future behavior and can remove auditability of prior reasoning or user instructions.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 166)May include surrounding context.

md
| .renal/filter-state.json | 当前过滤状态 |
| .dream/task-state.json | 参考:进行中任务数 |
| .dream/topic-state.json | 参考:活跃话题数 |
| skills/apollo-renal/SKILL.md | 本文档 |

## v1 实施规格

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This shell script's user-facing comments and all runtime output are in Chinese, including status messages and recommendations, with no indication that language selection is configurable or that the skill is intentionally limited to Chinese-speaking users. That creates a natural-language locale policy issue because the skill effectively enforces a single language without user opt-in.

Content

No source excerpt is available for this finding.