[SergeyMenshykh]

Motivation & Context

Tools can declare a FunctionInvocationContext parameter to access the current invocation. This PR brings the same opt-in pattern to skill scripts.

Description & Review Guide

• What are the major changes?
  • SkillsProvider passes the invocation through SkillScript.run(...). A custom run() opts in by declaring a parameter annotated with FunctionInvocationContext; run() implementations without that annotation keep receiving host values as keyword arguments.
  • Inline script callables and file script runners can declare one FunctionInvocationContext parameter. The framework injects it, and it is hidden from the script schema. Runners must give it a default value.
  • InlineSkillScript.run and FileSkillScript.run gain an optional keyword-only context parameter; explicit application values passed as context are still forwarded unchanged.
  • Samples and docs updated.
• What is the impact of these changes? Opt-in only. Scripts, runners, and run() overrides without a context parameter behave as before.
• What do you want reviewers to focus on? How the context parameter is detected on run(), callbacks, and runners.

Related Issue

N/A

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR is linked to an issue and there is no other open PR for this issue (see Related Issue above).
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

Copilot review overview

🟡 Changes recommended

Context detection incorrectly treats ordinary generic annotations containing FunctionInvocationContext as injectable parameters.

Get a fresh assessment by requesting another Copilot review.

Review effort: Balanced
Findings: 1

Open (1)

• Avoid misclassifying container annotations as injected context · New

What changed in this PR

Adds opt-in FunctionInvocationContext injection for skill scripts while preserving legacy execution behavior.

Changes:

• Adds context-aware inline and file script execution.
• Updates samples and documentation.
• Adds extensive sync, async, streaming, approval, and compatibility tests.

File	Description
python/​packages/​core/​agent_framework/​_skills.py	Implements context detection and injection.
python/​packages/​core/​tests/​core/​test_skills.py	Adds comprehensive regression coverage.
python/​packages/​core/​AGENTS.md	Documents context-aware scripts and runners.
python/​samples/​02-agents/​skills/​README.md	Documents runtime context usage.
python/​samples/​02-agents/​skills/​code_defined_skill/​README.md	Explains context in code-defined skills.
python/​samples/​02-agents/​skills/​code_defined_skill/​code_defined_skill.py	Demonstrates inline context injection.
python/​samples/​02-agents/​skills/​class_based_skill/​README.md	Explains context in class-based skills.
python/​samples/​02-agents/​skills/​class_based_skill/​class_based_skill.py	Demonstrates method context injection.

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

MAF Automated Review — Iteration 1

Result: Findings reported
Scope: full PR (1 commit(s)): 482299dfa131
Model: gpt-5.6-sol-fast

Overview

This PR adds opt-in context injection for inline skill callbacks and file runners while preserving legacy run overrides, path revalidation, and approval behavior. Tests cover supported annotation forms, callable runner variants, context isolation, and reserved-argument rejection. However, all-or-nothing annotation resolution leaves an uncovered postponed-annotation case that silently treats a declared context parameter as model input, violating the intended trust boundary.

Reviewed the supplied pull-request change set across correctness, security/reliability, architecture, and failure behavior.
1 verified finding remained after source verification (1 medium) across 1 file. Details are attached to the affected lines below.

Affected areas: python/packages/core/agent_framework/_skills.py

[github-actions]

MAF Automated Review — Iteration 2

Result: No findings
Scope: 19 net-new commit(s): 30b9b8e06766, 622737258c73, 7fb41d673c97, 4f4e72284c9c, 2055e666e1ac, 62337887116c, b20927b921d4, 2191c60c5e7d, 717557a3791a, 91fd47035684, f83a892f9f0e, 0843a9c0537b, e2c6df28daa9, 72272613d0c0, 5ee3e87d6768, 8ff549d3f721, 024dd9908bc1, 2c46deb91e70, 19797cba63be
Model: gpt-5.6-sol-fast

Overview

The PR adds opt-in invocation-context injection while preserving legacy script and runner dispatch. Registration-time validation, reserved-argument checks, approval ordering, path validation, and broad sync/async and concurrency coverage constrain the changed behavior. No new Critical, High, or Medium defect survived reconciliation against the implementation, tests, current incremental diff, and existing feedback.

Reviewed the supplied incremental change set across correctness, security/reliability, architecture, and failure behavior.
No publishable findings remained after source verification for this scope.