feat(agents): project-local agent overrides

loadAgents merges <cwd>/agents/*.md on top of the extension's baseline:
a repo agent wins on name collision and may add brand-new agents. The
unknown-agent error now names both searched directories.
This commit is contained in:
2026-08-09 16:45:30 -04:00
parent c605a709fb
commit cb8a88d0bd
4 changed files with 149 additions and 23 deletions

View File

@@ -137,7 +137,10 @@ registerCheck(def) ← checks/*.ts self-register on load
`cwd` (see `src/agent-runner.ts`), with the agent definition's system prompt
and tool allowlist applied. Agent definitions are plain editable markdown in
[`agents/`](./agents/) (frontmatter `name` + `allowedTools`, body = system
prompt) — tuning a sub-agent never needs TypeScript changes.
prompt) — tuning a sub-agent never needs TypeScript changes. A scanned
project can ship its own `agents/*.md` at its root: those are loaded as
overrides (repo agent wins on name collision), so teams can tune prompts or
add project-specific agents without touching the extension.
- **Run-state** is a single JSON file at `<cwd>/.pygienium/run-state.json`
(`src/run-state.ts`): per-check phase progress, captured findings/changes
text, and recon status. `/pygienium-status`, `/pygienium-resume`, and

View File

@@ -18,6 +18,7 @@
import { mkdir, writeFile } from "node:fs/promises";
import { dirname, isAbsolute, join } from "node:path";
import type { AgentSessionEvent } from "@earendil-works/pi-coding-agent";
import { loadAgents, extensionRoot, type AgentDef } from "./agents.js";
import type { ToolCallEntry } from "./phases.js";
export interface AgentTaskOptions {
@@ -74,13 +75,19 @@ export function resetAgentRunner(): void {
export async function defaultAgentRunner(
opts: AgentTaskOptions,
): Promise<AgentRunResult> {
const agent =
opts.agent ?? (await loadAgents({ cwd: opts.cwd })).get(opts.agentName);
const agents = await loadAgents({ cwd: opts.cwd });
const agent = opts.agent ?? agents.get(opts.agentName);
if (!agent) {
const names = [...agents.keys()];
return {
ok: false,
text: "",
error: `Unknown agent definition: "${opts.agentName}". Add agents/${opts.agentName}.md.`,
toolCalls: [],
error:
`Unknown agent definition: "${opts.agentName}". ` +
`Available agents: ${names.length > 0 ? names.join(", ") : "(none loaded — check agents/ directories exist)"}. ` +
`Searched ${extensionRoot()}/agents/ (extension) and ${opts.cwd}/agents/ (project-local). ` +
`Add agents/${opts.agentName}.md to either location.`,
};
}

View File

@@ -117,34 +117,77 @@ function asString(value: unknown): string | undefined {
}
/**
* Load every `agents/*.md` definition from the extension's `agents/` dir.
* Returns a map keyed by agent name. Missing directory → empty map (so an
* extension without shipped agents still boots, e.g. in tests).
* Load agent definitions: the extension's `agents/*.md` baseline, plus any
* repo-local `agents/*.md` at `<cwd>/agents/` (when `cwd` is given). Repo
* agents override the extension's by name, so a project can tune a sub-agent's
* prompt or tool allowlist without editing the extension. Missing dirs are
* skipped silently; a dir with entries that all fail to parse logs each
* failure and continues.
*/
export async function loadAgents(_opts?: {
export async function loadAgents(opts?: {
cwd?: string;
}): Promise<Map<string, AgentDef>> {
const dir = join(extensionRoot(), "agents");
const extRoot = extensionRoot();
const result = new Map<string, AgentDef>();
// Extension-shipped agents are the baseline.
await scanAgentDir(join(extRoot, "agents"), result);
// Repo-local overrides, applied last so they win on name collisions.
if (opts?.cwd) {
await scanAgentDir(join(opts.cwd, "agents"), result, true);
}
return result;
}
/**
* Load every `agents/*.md` in `dir` into `result` (later dirs win on name
* collisions). `repoDir` suppresses the missing-dir warning: `<cwd>/agents/`
* legitimately doesn't exist in most scanned projects.
*/
async function scanAgentDir(
dir: string,
result: Map<string, AgentDef>,
repoDir = false,
): Promise<void> {
let entries: string[];
try {
entries = await readdir(dir);
} catch {
return result;
} catch (err) {
if (!repoDir) {
console.error(
`[pygienium] agent loading: could not read agents dir at ${dir}: ${err instanceof Error ? err.message : String(err)}`,
);
}
return;
}
for (const entry of entries) {
if (!entry.endsWith(".md")) continue;
const sourcePath = join(dir, entry);
const raw = await readFile(sourcePath, "utf8");
const { frontmatter, body } = parseFrontmatter(raw);
const name = asString(frontmatter.name) ?? entry.slice(0, -".md".length);
const allowedTools = asStringList(frontmatter.allowedTools);
result.set(name, {
name,
systemPrompt: body,
allowedTools,
sourcePath,
});
try {
const raw = await readFile(sourcePath, "utf8");
const { frontmatter, body } = parseFrontmatter(raw);
const name = asString(frontmatter.name) ?? entry.slice(0, -".md".length);
const allowedTools = asStringList(frontmatter.allowedTools);
result.set(name, {
name,
systemPrompt: body,
allowedTools,
sourcePath,
});
} catch (err) {
console.error(
`[pygienium] agent loading: failed to load ${entry} from ${dir}: ${err instanceof Error ? err.message : String(err)}`,
);
// Continue loading other agents even if one fails.
}
}
if (!repoDir && result.size === 0 && entries.length > 0) {
console.error(
`[pygienium] agent loading: found ${entries.length} entries in ${dir} but loaded 0 agents`,
);
}
return result;
}

View File

@@ -2,8 +2,29 @@
* agents.test.ts — markdown agent-definition loader.
*/
import { describe, expect, it } from "bun:test";
import { mkdtemp, mkdir, rm, writeFile } from "node:fs/promises";
import { tmpdir } from "node:os";
import { join } from "node:path";
import { extensionRoot, loadAgents } from "../src/agents.js";
/** Build an agent markdown file with frontmatter + body. */
async function writeAgent(
dir: string,
file: string,
name: string,
body: string,
tools?: string[],
): Promise<string> {
const lines = ["---", `name: ${name}`];
if (tools) {
lines.push("allowedTools:");
for (const t of tools) lines.push(` - ${t}`);
}
lines.push("---", body, "");
await writeFile(join(dir, file), lines.join("\n"), "utf8");
return join(dir, file);
}
describe("loadAgents", () => {
it("loads scanner.md and fixer.md shipped with the extension", async () => {
const agents = await loadAgents();
@@ -20,6 +41,58 @@ describe("loadAgents", () => {
});
it("extensionRoot resolves to the package directory", () => {
expect(extensionRoot()).toMatch(/pygienium$/);
expect(extensionRoot()).toMatch(/(pygenium|pygienium)$/);
});
it("project-local agents override the extension's by name", async () => {
const cwd = await mkdtemp(join(tmpdir(), "pygium-agents-"));
try {
await mkdir(join(cwd, "agents"));
const path = await writeAgent(
join(cwd, "agents"),
"scanner.md",
"scanner",
"Project-tuned scanner prompt.",
["read", "grep", "find", "edit"],
);
const agents = await loadAgents({ cwd });
const scanner = agents.get("scanner");
expect(scanner?.systemPrompt).toContain("Project-tuned");
expect(scanner?.allowedTools).toContain("edit"); // repo override widens tools
expect(scanner?.sourcePath).toBe(path);
// Extension baseline is retained for agents the repo doesn't override.
expect(agents.get("fixer")).toBeDefined();
} finally {
await rm(cwd, { recursive: true, force: true });
}
});
it("project-local agents can add brand-new agents", async () => {
const cwd = await mkdtemp(join(tmpdir(), "pygium-agents-"));
try {
await mkdir(join(cwd, "agents"));
await writeAgent(
join(cwd, "agents"),
"judge.md",
"judge",
"Scoring judge.",
);
const agents = await loadAgents({ cwd });
expect(agents.has("judge")).toBe(true);
expect(agents.get("judge")?.systemPrompt).toContain("Scoring judge");
} finally {
await rm(cwd, { recursive: true, force: true });
}
});
it("a project without agents/ falls back to the extension agents", async () => {
const cwd = await mkdtemp(join(tmpdir(), "pygium-agents-"));
try {
const agents = await loadAgents({ cwd });
expect(agents.has("scanner")).toBe(true);
expect(agents.has("fixer")).toBe(true);
} finally {
await rm(cwd, { recursive: true, force: true });
}
});
});