feat(eval): EvalRunner trait, aikit RunOptions timeout/cwd, trace parity

src/cli/cli.rs

@@ -5,7 +5,7 @@ use fastskill::FastSkillService;
 use std::sync::Arc;
 
 use crate::cli::commands::{
-    add, analyze, auth, disable, init, install, list, marketplace, package, publish, read,
+    add, analyze, auth, disable, eval, init, install, list, marketplace, package, publish, read,
     registry, reindex, remove, repos, search, serve, show, sources, sync, update, version,
     Commands,
 };
@@ -109,6 +109,10 @@ impl Cli {
             return auth::execute_auth(args).await;
         }
 
+        if let Some(Commands::Eval(args)) = self.command {
+            return eval::execute_eval(args).await;
+        }
+
         // For other commands, we need to restore self.command, so we need a different approach
         // Actually, if we get here, command was None or not Init/Repository/Package/Publish/RegistryIndex
         let command = self.command;
@@ -155,6 +159,7 @@ impl Cli {
             | Some(Commands::Repos(_))
             | Some(Commands::Marketplace(_))
             | Some(Commands::Auth(_))
+            | Some(Commands::Eval(_))
             | Some(Commands::Version(_)) => unreachable!("Handled above"),
             None => {
                 // No subcommand - treat as skill ID for read command

src/cli/commands/eval/mod.rs

@@ -0,0 +1,62 @@
+//! Eval command group for skill quality assurance
+
+pub mod report;
+pub mod run;
+pub mod score;
+pub mod validate;
+
+use crate::cli::error::CliResult;
+use clap::{Args, Subcommand};
+
+/// Eval command group
+#[derive(Debug, Args)]
+#[command(
+    about = "Evaluation commands for skill quality assurance",
+    after_help = "Examples:\n  fastskill eval validate\n  fastskill eval run --agent codex --output-dir /tmp/evals"
+)]
+pub struct EvalCommand {
+    #[command(subcommand)]
+    pub command: EvalSubcommand,
+}
+
+/// Eval subcommands
+#[derive(Debug, Subcommand)]
+pub enum EvalSubcommand {
+    /// Validate eval configuration and files
+    #[command(
+        about = "Validate eval configuration and files",
+        after_help = "Examples:\n  fastskill eval validate\n  fastskill eval validate --agent codex"
+    )]
+    Validate(validate::ValidateArgs),
+
+    /// Run eval cases against an agent
+    #[command(
+        about = "Run eval cases against an agent",
+        after_help = "Examples:\n  fastskill eval run --agent codex --output-dir /tmp/evals"
+    )]
+    Run(run::RunArgs),
+
+    /// Show a report for a completed eval run
+    #[command(
+        about = "Show a report for a completed eval run",
+        after_help = "Examples:\n  fastskill eval report --run-dir /tmp/evals/2026-04-01T14-32-10Z"
+    )]
+    Report(report::ReportArgs),
+
+    /// Re-score saved eval artifacts without running the agent again
+    #[command(
+        about = "Re-score saved eval artifacts without running the agent again",
+        after_help = "Examples:\n  fastskill eval score --run-dir /tmp/evals/2026-04-01T14-32-10Z"
+    )]
+    Score(score::ScoreArgs),
+}
+
+/// Execute the eval command group
+pub async fn execute_eval(args: EvalCommand) -> CliResult<()> {
+    match args.command {
+        EvalSubcommand::Validate(args) => validate::execute_validate(args).await,
+        EvalSubcommand::Run(args) => run::execute_run(args).await,
+        EvalSubcommand::Report(args) => report::execute_report(args).await,
+        EvalSubcommand::Score(args) => score::execute_score(args).await,
+    }
+}

src/cli/commands/eval/report.rs

@@ -0,0 +1,80 @@
+//! Eval report subcommand - artifact summary and formatting
+
+use crate::cli::commands::common::validate_format_args;
+use crate::cli::error::{CliError, CliResult};
+use clap::Args;
+use fastskill::eval::artifacts::read_summary;
+use fastskill::OutputFormat;
+use std::path::PathBuf;
+
+/// Arguments for `fastskill eval report`
+#[derive(Debug, Args)]
+#[command(
+    about = "Show a report for a completed eval run",
+    after_help = "Examples:\n  fastskill eval report --run-dir /tmp/evals/2026-04-01T14-32-10Z\n  fastskill eval report --run-dir ./evals/2026-04-01T14-32-10Z --json"
+)]
+pub struct ReportArgs {
+    /// Path to the specific run directory
+    #[arg(long, required = true)]
+    pub run_dir: PathBuf,
+
+    /// Output format: table, json, grid, xml (default: table)
+    #[arg(long, value_enum, help = "Output format: table, json, grid, xml")]
+    pub format: Option<OutputFormat>,
+
+    /// Shorthand for --format json
+    #[arg(long, help = "Shorthand for --format json")]
+    pub json: bool,
+}
+
+/// Execute the `eval report` command
+pub async fn execute_report(args: ReportArgs) -> CliResult<()> {
+    let format = validate_format_args(&args.format, args.json)?;
+    let use_json = format == OutputFormat::Json;
+
+    if !args.run_dir.exists() {
+        return Err(CliError::Config(format!(
+            "EVAL_ARTIFACTS_CORRUPT: Run directory does not exist: {}",
+            args.run_dir.display()
+        )));
+    }
+
+    let summary = read_summary(&args.run_dir).map_err(|e| {
+        CliError::Config(format!(
+            "EVAL_ARTIFACTS_CORRUPT: Failed to read summary.json: {}",
+            e
+        ))
+    })?;
+
+    if use_json {
+        println!(
+            "{}",
+            serde_json::to_string_pretty(&summary).unwrap_or_default()
+        );
+    } else {
+        println!("Eval Report");
+        println!("  run_dir: {}", args.run_dir.display());
+        println!("  agent: {}", summary.agent);
+        if let Some(ref model) = summary.model {
+            println!("  model: {}", model);
+        }
+        println!(
+            "  result: {}",
+            if summary.suite_pass {
+                "PASSED"
+            } else {
+                "FAILED"
+            }
+        );
+        println!("  cases: {}/{} passed", summary.passed, summary.total_cases);
+
+        if !summary.cases.is_empty() {
+            println!("\nCase Results:");
+            for case in &summary.cases {
+                println!("  [{}] {}", case.status, case.id);
+            }
+        }
+    }
+
+    Ok(())
+}