Skills autonomous-ci
📦
autonomous-ci
Content revision r1 High Risk ⚙️ External commands🌐 Network access
Verify CI Before Completion
Agents can report work as done before local tests and CI prove it. This skill defines a verification workflow that checks tests, GitHub Actions, failures, and final evidence.
Supports: Claude Codex Code(CC)
Stop for confirmation before installing.
Review the plan and obtain explicit user consent before changing files.
Install with my Agent
Copy this request to your Agent. It includes the canonical Skill page and manifest.
Agent request
Review the Skillstore skill "autonomous-ci" from https://skillstore.io/skills/ancplua-autonomous-ci.md and its manifest at https://skillstore.io/api/skills/ancplua-autonomous-ci/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.Your Agent should still show its plan and request any confirmation required by the security policy.
Agent-readable resources
Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.
Test it
Using "autonomous-ci". Verify the latest bug fix before I open a pull request.
Expected outcome:
Local tests passed. GitHub Actions completed successfully for build, test, and lint. The response includes workflow links and the verified commit.
Using "autonomous-ci". CI failed after my push. Find out why.
Expected outcome:
The failed workflow log points to a test assertion in the payment module. The report names the failing check and recommends the next fix.
Using "autonomous-ci". Can we call this task complete?
Expected outcome:
Not yet. Local tests passed, but one remote workflow is still running. Completion should wait for the final CI result.
Security Audit
High Riskv8 • 7/5/2026 Open versioned report
Several Markdown fence detections are false positives, but the skill intentionally directs agents to run local build and test commands, use GitHub CLI, and collect CI results. The highest risk is the mandatory commit-and-push workflow, which can mutate local and remote repository state without a separate consent gate. No credential exfiltration endpoint or explicit prompt injection to bypass security review was found.
1
Files scanned
273
Lines analyzed
18
Review items
0
False positives ignored
Confirmed security concerns (2)
High
Autonomous Git Push Requirement
The skill requires agents to commit and push code before claiming completion. This can publish local changes or unrelated files unless explicit user approval is required.
The workflow explicitly includes COMMIT AND PUSH and lists git add ., git commit, and git push commands. The side effect is clear and directly stated.
Medium
Overbroad Mandatory Agent Instructions
The skill says it is mandatory before many responses and states there are no exceptions. This can pressure the agent to follow the skill even when the user has not requested CI or remote writes.
The mandatory and no-exceptions language is explicit, but it does not contain a direct system override or instruction to bypass security review.
Capability review items (18)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
Ruby/shell backtick execution
```
The nearby section introduces a required commit and push step after tests pass. That workflow can mutate local git state and publish changes to a remote repository.
High
Ruby/shell backtick execution
```bash
The code block contains git add ., git commit, and git push. Staging all files and pushing to a remote repository is a side-effecting action that requires explicit user control.
Medium
Ruby/shell backtick execution
```text
This section directs the agent to run local verification, commit and push, then monitor remote CI before reporting completion. The triple backticks are harmless, but the workflow instruction creates external command and remote repository risk.
Medium
Ruby/shell backtick execution
```bash
The Bash example tells the agent to run dotnet, npm, pytest, and go test commands. These are legitimate CI tasks, but they execute project-controlled code and scripts.
Medium
Ruby/shell backtick execution
```
The prose says the agent will auto-detect the project type and run the appropriate command, then use GitHub CLI. This is an instruction to execute local and network-capable tools.
Medium
Ruby/shell backtick execution
```bash
The code block invokes gh run list, gh run watch, and gh run view against GitHub CI. These external commands contact GitHub and rely on the user's authenticated CLI context.
Medium
Ruby/shell backtick execution
```bash
The Bash block instructs the agent to run build and test commands for .NET, Node.js, and Python projects. These commands can execute repository-defined scripts and test code.
Medium
Ruby/shell backtick execution
```
The cited span introduces blocking CI monitoring and opens the shell example that follows. It is weaker evidence than the command lines, but it still directs execution of external CI tooling.
Medium
Ruby/shell backtick execution
```bash
The Bash block runs git and gh commands, stores command output, watches a run, and downloads failed logs. This is real shell execution and network-backed CI access.
Medium
Ruby/shell backtick execution
```
The failure-handling section tells the agent to download CI failure logs with gh. This is an authenticated external command that may retrieve repository log data.
Medium
Ruby/shell backtick execution
1. Download failure logs: `gh run view <run-id> --log-failed`
The inline command gh run view --log-failed is explicitly presented as the first failure-handling step. It accesses GitHub CI logs through the authenticated CLI.
Medium
Ruby/shell backtick execution
```bash
The .NET example tells the agent to run dotnet test. It is a legitimate verification command, but it still executes project code on the user's machine.
Medium
Ruby/shell backtick execution
```bash
The code block invokes gh run list with a commit selector and returns workflow metadata. This is an authenticated external command to GitHub.
Medium
Ruby/shell backtick execution
```bash
The Bash example runs gh run watch to wait on a GitHub Actions run. This is an external command that contacts GitHub and blocks on remote CI state.
Medium
Ruby/shell backtick execution
```
The success criteria require code to be committed and pushed, CI workflows to be monitored, and URLs or logs to be collected. This confirms the intended external command workflow.
Medium
Shell command substitution
COMMIT_SHA=$(git rev-parse HEAD)
The command substitution runs git rev-parse HEAD in a shell. The command is fixed and low injection risk, but it is still shell execution as part of the workflow.
Medium
Shell command substitution
RUN_ID=$(gh run list --commit "$COMMIT_SHA" --limit 1 --json databaseId -q '.[0].databaseId')
The command substitution runs gh run list and captures a GitHub Actions run ID. Quoting reduces injection risk, but the command contacts GitHub using authenticated state.
Medium
Shell command substitution
CONCLUSION=$(gh run view "$RUN_ID" --json conclusion -q .conclusion)
The command substitution runs gh run view to retrieve a CI conclusion. It is hardcoded and quoted, but still uses an external command and GitHub network access.
Risk Factors
⚙️ External commands (25)
SKILL.md:28-53 SKILL.md:53-63 SKILL.md:63-75 SKILL.md:75-83 SKILL.md:83-90 SKILL.md:90-109 SKILL.md:109-123 SKILL.md:123-127 SKILL.md:127-132 SKILL.md:132-136 SKILL.md:136-155 SKILL.md:155-161 SKILL.md:161-181 SKILL.md:181-187 SKILL.md:187-193 SKILL.md:193-201 SKILL.md:201-207 SKILL.md:207-213 SKILL.md:213-229 SKILL.md:229-245 SKILL.md:245-258 SKILL.md:138 SKILL.md:141 SKILL.md:147 SKILL.md:136-155
🌐 Network access (3)
Audited by: codex View Audit History →
Share & cite this report
Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.
Copy report link
https://skillstore.io/skills/ancplua-autonomous-ci/audits/8?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown badge
[](https://skillstore.io/skills/ancplua-autonomous-ci?utm_source=security_passport_badge)HTML badge
<a href="https://skillstore.io/skills/ancplua-autonomous-ci?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/ancplua-autonomous-ci/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Embed card
<iframe src="https://skillstore.io/embed/skills/ancplua-autonomous-ci.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>Academic citations (APA · BibTeX · CFF)
APA citation
ANcpLua. (2026). autonomous-ci security audit report (audit version 8) [Author version unspecified]. Skillstore. https://skillstore.io/skills/ancplua-autonomous-ci/audits/8BibTeX citation
@techreport{ancplua-ancplua-autonomous-ci-2026,
author = {ANcpLua},
title = {autonomous-ci security audit report (audit version 8)},
institution = {Skillstore},
year = {2026},
number = {8},
url = {https://skillstore.io/skills/ancplua-autonomous-ci/audits/8},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "autonomous-ci security audit report (audit version 8)"
version: "unspecified"
type: report
authors:
- name: "ANcpLua"
date-released: "2026-07-05"
url: "https://skillstore.io/skills/ancplua-autonomous-ci/audits/8"
identifiers:
- type: other
value: "skillstore:ancplua-autonomous-ci:audit:8"
description: "Skillstore immutable audit report identifier"
Skillstore Score
Why this score Evidence Confidence: Medium 55
Architecture
85
Maintainability
87
Content
71
Community
83
Spec Compliance
What You Can Build
Verify a code fix before handoff
Run local tests, monitor GitHub Actions, and report completion only after evidence is available.
Gate pull request readiness
Check that build, test, and lint workflows pass before asking for review or merging work.
Investigate CI failures
Download failed workflow logs, identify likely causes, apply fixes, and repeat the verification loop.
Try These Prompts
Verify my change
Use the autonomous CI workflow for my latest change. Run local verification, then summarize what passed and what still needs attention.
Check GitHub Actions
Verify this branch with local tests and GitHub Actions. Ask before any commit or push, and include workflow links in the result.
Debug failed CI
Use the CI failure protocol. Inspect failed workflow logs, explain the likely root cause, fix it, and rerun local verification before reporting status.
Apply a release gate
Apply the full verification gate for this release branch. Confirm local tests, all required workflows, failure evidence, and final completion evidence.
Best Practices
- Ask for explicit permission before staging, committing, or pushing changes.
- Run the smallest useful local verification first, then confirm all required remote workflows.
- Include concrete evidence such as test counts, workflow names, conclusions, links, and commit identifiers.
Avoid
- Do not claim success based only on confidence or local reasoning.
- Do not ignore a failed or still-running CI workflow.
- Do not push code automatically when the user only asked for analysis or review.
Frequently Asked Questions
Does this skill run CI by itself?
It provides instructions for an agent to run local commands and monitor GitHub Actions. Tool execution still depends on the host agent.
Which CI service does it support?
The examples focus on GitHub Actions through the GitHub CLI. Other CI systems require manual changes to the workflow.
Can it be used without pushing code?
Yes, use it for local verification and read-only CI checks. Remote writes should require explicit user approval.
What project types are covered?
The examples cover .NET, Node.js, Python, and Go test commands. Other stacks need project-specific commands.
What evidence should the final response include?
A good report includes local test status, workflow names, CI conclusions, links or logs, commit identity, and remaining risks.
Is this safe for untrusted repositories?
Use caution. Test scripts and build hooks can execute code from the repository, so review commands before running them.