Trait HarnessAdapter 

pub trait HarnessAdapter {
    // Required methods
    fn label(&self) -> &'static str;
    fn skills_dir(&self, repo_root: &Path) -> PathBuf;
    fn rewrites_frontmatter_name(&self) -> bool;
    fn advertises_staged_slug_name(&self) -> bool;
    fn render_available_skills_block(&self, skills: &[AvailableSkill]) -> String;
    fn skill_surface_phrase(&self) -> &'static str;
    fn skill_unresolved_phrase(&self) -> &'static str;
    fn plan_mode_profile(&self) -> &'static str;
    fn parse_transcript(&self, path: &Path) -> Result<Vec<ToolInvocation>>;
    fn parse_transcript_full(&self, path: &Path) -> Result<TranscriptSummary>;
    fn install_guard(
        &self,
        stage_root: &Path,
        guard_exe: &Path,
        ttl: Option<Duration>,
    ) -> Result<PathBuf>;

    // Provided methods
    fn render_plan_mode_context(&self, profile_text: &str) -> String { ... }
    fn runbook_template(&self) -> &'static str { ... }
    fn cli_events_filename(&self) -> Option<&'static str> { ... }
    fn cli_model_flag(&self) -> Option<&'static str> { ... }
    fn cli_next_steps(&self, _ctx: CliDispatchContext<'_>) -> String { ... }
    fn cli_manifest_section(
        &self,
        _ctx: CliManifestContext<'_>,
    ) -> Option<Vec<String>> { ... }
    fn cli_judge_next_steps(&self, _ctx: CliJudgeContext) -> Option<String> { ... }
    fn parse_cli_events(&self, path: &Path) -> Result<Vec<ToolInvocation>> { ... }
    fn parse_cli_events_full(&self, path: &Path) -> Result<TranscriptSummary> { ... }
    fn guard_armed_message(&self) -> Option<&'static str> { ... }
}

The behavior that varies by harness. Generic run-mode code depends on this trait, never on a concrete harness variant.

Required Methods

fn label(&self) -> &'static str

The kebab-case identifier used in CLI flags, dispatch.json, and the staged conditions.json.

fn skills_dir(&self, repo_root: &Path) -> PathBuf

The project-local directory staged skills live under for this harness.

fn rewrites_frontmatter_name(&self) -> bool

Whether a staged skill’s frontmatter name: is rewritten to its slug so the harness’s repo-local discovery resolves the staged copy.

fn advertises_staged_slug_name(&self) -> bool

Whether the skill-under-test is advertised in the available-skills block under its staged slug (vs. its natural name). True for Codex, whose repo-local discovery keys on the rewritten frontmatter name. (OpenCode also rewrites the frontmatter to the slug yet still advertises the natural name — a known inconsistency tracked for a separate fix.)

fn render_available_skills_block(&self, skills: &[AvailableSkill]) -> String

Render the discoverable skills the way this harness natively surfaces them (e.g. Claude Code’s Skill-tool list, Codex’s ## Skills, OpenCode’s <available_skills> XML).

fn skill_surface_phrase(&self) -> &'static str

How a staged skill is described as discoverable in the neutral slug-disambiguation line (e.g. “via the Skill tool”).

fn skill_unresolved_phrase(&self) -> &'static str

The lead-in for the fallback “read the skill from <path>” instruction when the staged identifier can’t be resolved.

fn plan_mode_profile(&self) -> &'static str

The verbatim plan-mode procedure profile bundled for this harness.

fn parse_transcript(&self, path: &Path) -> Result<Vec<ToolInvocation>>

Parse a persisted transcript into its ordered tool invocations.

fn parse_transcript_full(&self, path: &Path) -> Result<TranscriptSummary>

Parse a persisted transcript into the full summary: tool invocations, deduped token usage, duration, and final message text.

fn install_guard( &self, stage_root: &Path, guard_exe: &Path, ttl: Option<Duration>, ) -> Result<PathBuf>

Arm the write guard using this harness’s native pre-tool hook surface, returning the staged marker path. The guard’s allowed roots are derived from stage_root (the isolated env / agent cwd), so it bounds the agent to the same env boundary that isolates its reads.

Provided Methods

fn render_plan_mode_context(&self, profile_text: &str) -> String

Wrap a plan-mode profile as a <system-reminder> operating-context layer. The default usually suffices.

fn runbook_template(&self) -> &'static str

The interactive (agent-followed) RUNBOOK.md template a harness uses under InSession dispatch, carrying {{TOKEN}} placeholders the run fills. The default is the shared headless template (harmless for the Cli-only harnesses that never read it via this path); InSession harnesses override it. The Cli-dispatch runbook always uses HEADLESS_RUNBOOK_TEMPLATE, selected by mechanism in build_runbook.

fn cli_events_filename(&self) -> Option<&'static str>

For a Cli-dispatch harness, the filename (under a task’s outputs/ dir) its one-shot CLI writes the transcript to. None when the harness dispatches in-session (no local transcript) or has no Cli-mechanism transcript wired yet.

fn cli_model_flag(&self) -> Option<&'static str>

For a Cli-dispatch harness, the native model-selection flag accepted by the harness CLI. None means the adapter has no model-selection support wired yet.

fn cli_next_steps(&self, _ctx: CliDispatchContext<'_>) -> String

The Next: guidance printed after run for a Cli-dispatch harness: how to dispatch each task through this harness’s one-shot CLI and then ingest. Empty for in-session harnesses (their guidance is the mechanism’s, not the adapter’s).

fn cli_manifest_section( &self, _ctx: CliManifestContext<'_>, ) -> Option<Vec<String>>

Extra dispatch-manifest.md lines describing this harness’s Cli dispatch recipe (command template, parallel recipe, ingest note). None when the harness contributes no Cli-specific manifest section.

fn cli_judge_next_steps(&self, _ctx: CliJudgeContext) -> Option<String>

The post-grade / post-ingest judge dispatch guidance for a Cli-dispatch harness. None leaves the generic in-session-style judge handoff in place.

fn parse_cli_events(&self, path: &Path) -> Result<Vec<ToolInvocation>>

Parse a Cli-mechanism events file (the harness CLI’s captured output) into ordered tool invocations. Defaults to parse_transcript: for Codex/OpenCode the on-disk parser already is the events parser, so the default is correct; Claude Code overrides it, because its parse_transcript is the in-session subagent parser while its Cli events are claude -p stream-json.

fn parse_cli_events_full(&self, path: &Path) -> Result<TranscriptSummary>

The full-summary counterpart of parse_cli_events.

fn guard_armed_message(&self) -> Option<&'static str>

The banner printed after --guard successfully arms, describing the harness’s native hook surface and how to remove it. Harness-specific text, so it lives here rather than in generic run code. None for a harness with no write guard (its install_guard errors), in which case no banner is printed.

Dyn Compatibility

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors

impl HarnessAdapter for ClaudeCodeAdapter

impl HarnessAdapter for CodexAdapter

impl HarnessAdapter for OpenCodeAdapter