Most static alerts misread Markdown syntax, reference links, or read-only examples. Four network findings are confirmed because the templates prescribe state-changing requests and an applyable broad network policy. One semantic finding covers production-changing commands that lack immediate approval safeguards; no prompt injection or exfiltration was found.
The templates present rollback, scaling, database termination, feature-flag, and network-policy commands without requiring confirmation immediately before execution. Misapplied examples can disrupt services or broaden access.
The cited sections contain explicit mutating production commands. General validation guidance at the end reduces intent concerns but does not add per-action confirmation.
Capability review items (4)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
curl -X POST https://api.company.com/internal/feature-flags \
The example sends a POST request that enables a feature flag at an internal endpoint. If adapted and executed without approval, it can change production behavior.
The applyable NetworkPolicy example permits 0.0.0.0/0 except one subnet. Because policies are additive, this can unintentionally broaden production ingress.
The policy hardcodes 192.168.1.0/24 as the suspicious range. Copying the placeholder can block legitimate private traffic while failing to block the actual source.
All static findings are documentation artifacts in SKILL.md, including Markdown code fences, inline examples, sample URLs, sample CIDR values, and /dev/null usage. I found no executable scripts, prompt injection text, credential handling, hidden network behavior, or data exfiltration intent.
All static findings are Markdown documentation examples in SKILL.md, not executable skill code. The shell, SQL, URL, CIDR, and /dev/null matches are runbook placeholders or manual commands, with no evidence of automatic execution, exfiltration, or prompt injection.
Static analysis found many command, URL, device-file, IP, and weak-crypto patterns in SKILL.md. Review shows the command and network items are Markdown runbook examples, and the weak-crypto alerts are text-match false positives; the remaining risk is that users may copy disruptive operational commands without adapting them to their environment.
Operational Runbook Commands Require Review Before Use
The skill contains fenced Markdown examples for kubectl, psql, curl, and local scripts. These are not executed by the skill, but copying them into production can change deployments, terminate database sessions, alter feature flags, or run rollback scripts.
The commands are clearly present in documentation examples and include production-changing operations. Confidence is high that this is a real operator safety concern, but not malicious code execution.
Static false positives ignored (4)
These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.
Static Shell Execution Alerts Are Markdown Examples
The Ruby or shell backtick alerts correspond to Markdown fences, inline code, and example commands in SKILL.md. No executable Ruby code, shell wrapper, or automatic command invocation was found in the scanned file.
The flagged regions are visibly Markdown templates and fenced command examples. The repository file list contains only SKILL.md and no executable implementation file for automatic command execution.
The weak cryptographic algorithm alerts appear to match ordinary text such as frontmatter description fields and SQL ORDER BY DESC examples. No cryptographic API, cipher selection, or hashing implementation was found.
The cited lines are metadata or SQL sort clauses, not cryptographic code. There is no evidence of DES, MD5, SHA1, or another weak algorithm being used by the skill.
The skill includes dashboard links, API health checks, Prometheus queries, external references, 0.0.0.0/0, and a private CIDR in example runbooks. These are documentation placeholders or operational examples, not automatic outbound requests.
The lines are Markdown links, curl examples, and Kubernetes network policy examples. No code path sends data to these URLs automatically.
All 62 static findings are false positives from Markdown documentation. The skill contains only template documentation with code examples (bash, kubectl, SQL) in fenced code blocks. No executable code, no prompt injection attempts, and no security risks detected. Safe to publish.
1
Files scanned
398
Lines analyzed
0
Review items
0
False positives ignored
No confirmed security findings were recorded for this completed audit.