NetFabric/NetFabric.Numerics / .agents/skills/copilot-github-actions
copilot-github-actions
VerifiedAutomate GitHub Copilot CLI in GitHub Actions workflows. USE FOR: installing/running `copilot -p` in workflow steps, choosing GITHUB_TOKEN vs a personal access token (COPILOT_GITHUB_TOKEN) for authentication, setting the copilot-requests write permission, minimal-permission tool allowlisting (--allow-tool, --deny-tool, --no-ask-user, -s/--silent, --output-format=json), triggers (schedule, workflow_dispatch, push), the recommended GitHub Agentic Workflows (gh-aw) alternative to direct CLI invocation, and security/cost considerations (fork PR prompt-injection risk, billing attribution, cost centers). DO NOT USE FOR: embedding the Copilot agent runtime into your own application (use copilot-sdk instead); general GitHub Actions authoring unrelated to Copilot; the Copilot code review or PR features outside of CLI automation.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.8 KB |
| references/authentication.md | markdown | 2.8 KB |
| references/cli-invocation-patterns.md | markdown | 3.2 KB |
| references/gh-aw-vs-direct.md | markdown | 2.7 KB |
| references/security-and-cost.md | markdown | 2.3 KB |
references/security-and-cost.md
View on GitHub# Security and Cost
## Security considerations
Copilot CLI is agentic — it can read and modify repository contents. A compromised or misconfigured workflow can cause unintended changes, independent of which auth method you use.
To reduce risk:
* Prefer [GitHub Agentic Workflows](gh-aw-vs-direct.md) over raw `copilot` steps — they're designed with guardrails for automated use.
* Follow least privilege on `permissions:` — grant only `contents: read` (or whatever the task needs) plus `copilot-requests: write`; avoid `contents: write` unless the job actually commits/pushes.
* Scope tool access narrowly with `--allow-tool`/`--deny-tool` (see [cli-invocation-patterns.md](cli-invocation-patterns.md)) instead of `--allow-all`/`--yolo`.
* **Review triggers carefully.** Workflows that run on `pull_request` events from forks are at the highest risk: the PR author fully controls the diff/commit messages/issue text the CLI will read, making prompt injection possible if the CLI is invoked directly in that step. Prefer `pull_request_target` combined with strict guardrails, or `gh-aw`'s built-in fork protections, over a naive `pull_request` + direct CLI invocation.
* Invoking Copilot CLI directly in workflow steps gives it broad access to the workflow environment (secrets in scope, filesystem, network) — treat that step like any other privileged CI step.
## Billing and cost
| Auth method | Attribution |
|---|---|
| PAT | Billed to the PAT owner's individual Copilot seat/license |
| `GITHUB_TOKEN`, personally-owned repo | Billed to the repo owner's seat |
| `GITHUB_TOKEN`, org-owned repo | Metered directly to the organization (requires the org policy); **no per-user budget applies** |
Because org-attributed usage bypasses individual user budgets, manage spend at the org level instead:
* Configure **cost centers** to attribute and cap spend across groups of organizations.
* Monitor consumption via the organization's billing and usage dashboards.
## Fork PR risk in one line
If a workflow step runs `copilot -p "..."` against content an external contributor controls (a fork's PR title/body/diff) with write-capable tools allowed, that contributor can potentially steer the agent (prompt injection) into taking unwanted actions with the workflow's credentials — scope permissions and tools accordingly, or avoid direct CLI invocation on fork-triggered events entirely.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |