Security
Security and Hardening
by Arsalan Usmani·Posted Sep 14, 2026
Reviews code and configurations for security weaknesses and applies hardening best practices before they reach production.
Security and Hardening scans application code, dependencies, and configuration files for vulnerabilities, exposed secrets, and insecure defaults. It maps findings to OWASP LLM Top 10 and ISO 27001 controls, then proposes concrete fixes and hardening steps. Every run produces evidence logs so security teams can audit what was checked and what changed.
What it does
7 actions, each one checked against your rules and recorded before it runs.
scan_vulnerabilities
Scan source code for injection, XSS, and OWASP LLM Top 10 weakness patterns.
audit_dependencies
Check third-party packages against known CVE feeds and version floors.
detect_secrets
Identify hardcoded credentials, API keys, and tokens, and redact them from output.
review_configurations
Evaluate configuration files against hardening baselines and ISO 27001 controls.
map_compliance
Map each finding to OWASP LLM and ISO 27001 control references.
propose_fixes
Generate minimal code or configuration patches that remediate findings.
assess_risk
Assign severity and exploitability context to each confirmed issue.
How it runs
Every run follows the same governed path, from request to recorded result.
- 1
The skill receives a repository path or diff plus optional configuration files.
- 2
It scans source code and dependencies for known vulnerability patterns and CVEs.
- 3
It detects hardcoded secrets and redacts them from all outputs.
- 4
It reviews configurations against hardening baselines and ISO 27001 controls.
- 5
It maps each finding to OWASP LLM Top 10 categories and assigns a severity.
- 6
It drafts concrete fixes and guardrail recommendations.
- 7
It emits a structured findings report with evidence logs for audit.
Governance & audit
What gets captured and verified on every single run.
Scoped repository access
Repository access is scoped to the target paths supplied in the request. No writes occur outside an approved branch or patch artifact.
Human approval for code changes
Code fixes and dependency upgrades are proposed as diffs. A human reviewer approves before merge.
Secret redaction
Detected secrets are masked in findings, logs, and evidence. Raw secret values are never persisted.
Evidence retention
Scan logs and findings are retained for 180 days, then purged per policy.
Standards traceability
Findings are mapped to OWASP LLM Top 10 and ISO 27001 controls so audits can trace each issue to a standard.
Example in action
Pre-merge security review for a SaaS release
- 1
Submit the change set. The team submits the release branch and config files as the review target.
- 2
Scan and detect. The skill scans code and dependencies, flagging an SQL injection risk and an outdated library with a known CVE.
- 3
Catch exposed secrets. A hardcoded API token is found in a config file and redacted from all outputs.
- 4
Map to standards. Each finding is mapped to OWASP LLM and ISO 27001 controls with a severity rating.
- 5
Apply the fixes. The skill proposes patches for the query, the dependency, and the config, which the team merges after review.
Skill manifest
The exact instructions this skill ships with, shown in full. Nothing hidden.
---
id: security-and-hardening
version: 0.1.0
display_name: Security and Hardening
pack: community
category: it-cybersecurity
risk_tier: medium
tags:
- security
- hardening
- vulnerability
- code-review
- devsecops
primary_domain: it-cybersecurity
secondary_domains:
- engineering
capabilities:
- code-api-operations
- compliance-policy-checks
- reasoning-analysis
verticals:
- saas
skill_type: composite
mechanisms:
- validator
- cognitive
standards:
- owasp-llm
- iso-27001
input_schema:
type: object
required:
- target
properties:
target:
type: string
description: Repository path, diff, or code snippet to review for security weaknesses.
config_paths:
type: array
description: Optional list of configuration file paths to include in the hardening
review.
standards_profile:
type: string
description: Optional compliance profile to enforce, such as owasp-llm or iso-27001.
severity_threshold:
type: string
description: Minimum severity level to report, one of low, medium, high, or
critical.
output_schema:
type: object
required:
- findings
- report
properties:
findings:
type: array
description: List of identified security issues with severity, location, and
standards mapping.
fixes:
type: array
description: Proposed code or configuration changes that remediate each finding.
report:
type: string
description: Human-readable summary of the review, risks, and recommended hardening
steps.
passed:
type: boolean
description: True when no findings meet or exceed the severity threshold.
planner:
mode: dynamic
prompt_section: How to think about this task
max_steps: 15
max_duration_sec: 300
max_cost_usd: 0.2
model_tier: balanced
allowed_step_kinds:
- tool_call
- llm_call
- output
evidence:
required_types:
- log
capture_per_step: true
retention_days: 180
routing:
intent_examples:
- Review this pull request for security vulnerabilities before merge.
- Scan my repo for hardcoded secrets and insecure configurations.
- Harden this service against the OWASP LLM Top 10.
- Check this Terraform file against our ISO 27001 controls.
not_for:
- Run a live penetration test against production infrastructure.
- Respond to an active security incident or breach.
verifier_rules:
- Every finding includes a severity, a file location, and a mapping to at least one
standard.
- Any detected secret value is redacted in findings, fixes, and logs.
- The report lists only findings at or above the requested severity threshold.
---
## Overview
Security and Hardening reviews application code, dependencies, and configuration files for vulnerabilities and insecure defaults. It maps findings to OWASP LLM Top 10 and ISO 27001 controls, then proposes concrete fixes. Every run emits evidence logs for audit.
## How to think about this task
Start by understanding the target. Identify languages, frameworks, and configuration formats so the right checks apply. Scan code for injection, unsafe deserialization, broken access control, and prompt injection patterns from the OWASP LLM Top 10.
Check dependencies for known CVEs and flag outdated or unmaintained packages. Review configurations for insecure defaults such as open CORS, verbose error output, missing TLS, and overly broad permissions.
Detect hardcoded secrets and credentials. Redact every secret value before it reaches a finding, a fix, or a log. Map each confirmed issue to the relevant standard control and assign a severity.
Propose the smallest safe fix for each finding. Prefer configuration changes and dependency upgrades over rewrites. Summarize residual risk clearly so a reviewer can approve with confidence.
## Capabilities
- scan_vulnerabilities: Scan source code for injection, XSS, and OWASP LLM Top 10 weakness patterns.
- audit_dependencies: Check third-party packages against known CVE feeds and version floors.
- detect_secrets: Identify hardcoded credentials, API keys, and tokens, and redact them from output.
- review_configurations: Evaluate configuration files against hardening baselines and ISO 27001 controls.
- map_compliance: Map each finding to OWASP LLM and ISO 27001 control references.
- propose_fixes: Generate minimal code or configuration patches that remediate findings.
- assess_risk: Assign severity and exploitability context to each confirmed issue.
## Inputs
`target` is the repository path, diff, or code snippet under review. The skill scopes all reads to this target and never writes outside an approved branch or patch artifact.
`config_paths` is an optional list of configuration files to include, such as Dockerfile, Terraform, or environment templates. When omitted, the skill discovers common configuration files inside the target.
`standards_profile` selects the compliance lens for the review. Supported values include owasp-llm and iso-27001. When omitted, both standards apply by default.
`severity_threshold` filters reported findings. Set it to low, medium, high, or critical. Findings below the threshold are logged but excluded from the report.
## Outputs
`findings` is the structured list of confirmed issues. Each entry carries a severity, a file and line location, a standards mapping, and a short description. Secret values are always redacted.
`fixes` contains the proposed remediations as patches or configuration changes. Each fix references the finding it resolves and is safe to apply after human review.
`report` is the human-readable summary. It covers scope, findings by severity, applied standards, and recommended hardening steps for the team.
`passed` is a boolean gate. It is true only when no findings meet or exceed the severity threshold, which makes it suitable for CI checks.
## Tools
- connectors: git, github, dependency-feed
- operators: static-analysis, secret-scanner, config-linter
- primitives: llm_call, file_read, diff_write
## Guardrails
- Repository access is scoped to the target paths supplied in the request. No writes occur outside an approved branch or patch artifact.
- Code fixes and dependency upgrades are proposed as diffs. A human reviewer approves before merge.
- Detected secrets are masked in findings, logs, and evidence. Raw secret values are never persisted.
- Scan logs and findings are retained for 180 days, then purged per policy.
- Findings are mapped to OWASP LLM Top 10 and ISO 27001 controls so audits can trace each issue to a standard.
## Examples
A SaaS team points the skill at a release-candidate branch with `target` set to the repo path and `severity_threshold` set to high. The skill flags an SQL injection risk in a query builder, an outdated JSON parsing library with a known CVE, and a hardcoded API token in a config file. It redacts the token, proposes a parameterized query fix and a dependency bump, and maps each issue to OWASP and ISO 27001 controls. The report returns `passed: false`, and the team merges the fixes before release.Reviews
Reviews from teams using Security and Hardening.
No reviews yet
Be the first to share how Security and Hardening works for your team.
Discussion(0)
No comments yet. Be the first.