Checklists/AI Security/Coding Agent Security

Coding Agent Security Checklist

26 audit checks for coding agents covering prompt-to-sink execution, repository trust boundaries, tool misuse, secret exposure, approval bypass, and evidence integrity.

🚨 Coding Agent Threat Landscape

Coding agents collapse prompt handling, shell execution, repository access, package installation, and cloud credentials into a single runtime. The real failures often happen at the execution sink, not the prompt layer.

Prompt to shell a single untrusted issue, README, or ticket field can steer code execution if argument provenance is weak

Repository trust drift agents routinely treat repo files, tests, and config as trusted even when attacker-controlled

Approval surface global auto-approve and standing permissions turn one mistaken grant into persistent execution authority

Secret blast radius developer machines and CI contexts often expose package tokens, cloud creds, and SSH material to agent tooling

Evidence fragility logs, task history, and scratch files are often incomplete or mutable, weakening forensics

Supply chain overlap coding agents frequently install dependencies, run scripts, and modify CI, so AI misuse and package compromise compound

📄
Need the coding agent checklist as PDF? Message Carlos on Telegram.DM me on Telegram →
Showing 26 of 26 vulnerabilities
#1

Mixed-Trust Input Inventory

Critical

All prompt-adjacent inputs are classified by trust, including issues, PR text, READMEs, tests, package metadata, and generated files

#2

Prompt-to-Sink Mapping

Critical

Each untrusted input path is traced to execution sinks like shell, eval, package install, CI config, or file write

#3

Argument-Level Provenance

Critical

The system preserves provenance for command arguments, file paths, destinations, and generated code, not just for the top-level task

#4

Command Construction Safety

Critical

Shell commands are built with structured argument arrays or equivalent safe construction, never with raw concatenated attacker-controlled strings

#5

Dynamic Execution Controls

Critical

Generated code, eval helpers, notebook cells, migrations, and package scripts execute only inside constrained sandboxes

#6

Write-Scope Enforcement

High

Agents cannot write outside approved directories or mutate policy files like shell config, git hooks, IDE settings, and CI workflows without explicit re-approval

#7

Branch and Repo Target Validation

High

Agent actions are pinned to the intended repository, worktree, and branch, with safeguards against retargeting via poisoned instructions

#8

Untrusted Test Containment

High

Running repository tests, build scripts, or codegen is treated as executing untrusted code unless the repo is explicitly trusted

#9

Hook and Filter Neutralization

Critical

Git hooks, smudge/clean filters, aliases, and post-checkout behaviors are disabled or tightly controlled for agent sessions

#10

Credential Exposure Minimization

Critical

Agent runtimes receive only task-scoped credentials, not the operator's full shell environment or general-purpose cloud access

#11

Secret Use Mediation

High

Sensitive operations use mediated brokers or attested tooling where possible, instead of exposing raw reusable secrets to the agent

#12

Output Secret Scrubbing

High

Logs, patches, error traces, and generated docs are scrubbed for secrets before display, storage, or PR creation

#13

Approval Persistence Boundaries

Critical

Approvals expire appropriately and do not silently persist across sessions, workspaces, subagents, or unrelated action classes

#14

High-Risk Action Re-Authorization

Critical

Destructive or externally impactful actions require fresh authorization even if a broader permission was previously granted

#15

Delegated Worker Attenuation

High

Subagents and delegated workers receive reduced authority by default, not the full parent runtime

#16

Dependency Installation Review

High

Package installs, plugin adds, MCP connections, and remote setup scripts are gated by provenance checks and trust policy

#17

Build and CI Mutation Review

High

Changes to CI, release, deployment, or package publishing workflows receive enhanced scrutiny

#18

Plugin and Skill Provenance

High

Agent plugins, skills, and connector manifests are treated as executable policy artifacts and reviewed for hidden authority

#19

Network Egress Policy

High

Agents have explicit outbound network policy, with internal services, metadata endpoints, and credential brokers blocked by default

#20

Remote Destination Validation

High

PR comments, issue posts, webhooks, and other outbound destinations are validated against expected owners and repositories

#21

Cost and Token Abuse Controls

Medium

The platform detects runaway loops, oversized contexts, and expensive model routing triggered by malicious tasks or poisoned content

#22

Action Log Integrity

High

Every command, file write, approval decision, and external action is recorded with tamper-evident metadata

#23

Transcript and Artifact Retention

Medium

Conversation history, generated plans, and scratch artifacts are retained long enough for post-incident review

#24

Diff Reviewability

Medium

The system produces human-reviewable diffs and rationale for every material code change

#25

Rollback and Containment Readiness

Medium

Operators can revert agent changes, disable delegated workers, and revoke credentials quickly after suspicious behavior

#26

Owner-Harm Scenarios Tested

Critical

The audit explicitly tests whether the agent can betray its own operator through secret leak, destructive code change, or unauthorized remote action

Need a coding agent security audit?

Zealynx reviews the real authority surface of coding agents, from prompt-to-shell sinks and repo trust boundaries to approval semantics, secrets, and forensics.

Smart Contract Security Digest

Monthly exploit breakdowns, audit checklists, and DeFi security research — straight to your inbox

© 2026 Zealynx