thePR0M3TH3AN/bitvid
1
0 Fork 0
mirror of https://github.com/PR0M3TH3AN/bitvid.git synced 2026-10-27 07:52:24 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
de81198a68e27d5fed59eeb7bd8df6748b0670fc
bitvid/AGENTS.md
T

34 KiB
Raw Blame History

bitvid — AI Agent Guide

This guide tells AI agents how to keep bitvid aligned with the current product direction. Follow it whenever you touch code, content, or documentation inside this repository.

Important: Before starting any development or debugging, please review KNOWN_ISSUES.md. This file documents pre-existing test failures, environmental quirks, and architectural limitations that you should be aware of to avoid unnecessary investigation.

For any nostr related work, please review the nip documentation located in /docs/nips before beginning work.

Comment publishing guardrail: Session actors are limited to passive telemetry (e.g., view counters). Always require the logged-in Nostr signer when sending video comments through js/nostr/commentEvents.js. See the reminder in docs/nostr-event-schemas.md for context.

1. Release Channels: Main vs. Unstable

  • Main is the production track. Anything merged here must preserve today’s UX and magnet safety guarantees. Rollbacks should be painless: keep commits atomic, avoid destructive migrations, and leave feature flags in a known-good default (false unless product explicitly flips them).
  • Unstable is our experimentation lane. Gate risky behavior behind feature flags defined in js/constants.js and document the toggle/rollback plan in PR descriptions.
  • Emergency response: If a change regresses playback or breaks magnet parsing, revert immediately, call this out in the PR, and annotate the AGENTS.md changelog with remediation tips. Note: These regressions usually surface in dev logs and tests around playback/magnet helpers—reference them when triaging so future agents don’t repeat the mistake.

Browser logging policy

  • Route every browser log through js/utils/logger.js. Do not call console.* directly.
  • Use logger.dev.* for diagnostics that should disappear in production (flagged by IS_DEV_MODE/isDevMode). Use logger.user.* for operator-facing warnings and errors that must reach production consoles.
  • Before promoting a build, flip IS_DEV_MODE in config/instance-config.js to match the target environment so window.__BITVID_DEV_MODE__ and the logger channels behave correctly.
  • Review and follow docs/logging.md whenever you add new logging or touch inline scripts.

2. Mission: Hybrid Playback Strategy

  • Goal: Always deliver smooth playback while optimizing for the configured priority (hosted URL or P2P).
  • Configuration: DEFAULT_PLAYBACK_SOURCE in config/instance-config.js determines whether url or torrent is attempted first.
  • Runtime behavior: Call bitvidApp.playVideoWithFallback({ url, magnet }) (implemented in js/app/playbackCoordinator.js and bound to the app instance in js/app.js). This function attempts the preferred source first and falls back to the alternative if the primary fails or stalls. js/playbackUtils.js now provides magnet/session helper utilities rather than the playback entry point itself.
  • Content contract: bitvid posts (Nostr kind 30078) must include a title and at least one of url or magnet. Prefer to publish both along with optional thumbnail, description, and mode fields. When mirroring to NIP‑94 (kind 1063), copy the hosted URL and (optionally) the magnet so other clients can discover the same asset.

Styling & Theming Rules (token-first)

  • Single source of truth: All colors, spacing, radii, and typography decisions must flow through our design tokens; never reach for ad-hoc HEX/RGB values when a token exists.
  • Theme scopes: Tokens live at three levels—core (global primitives), theme (light/dark surface/background mappings), and component (contextual overrides). Additions must document their scope and consumers.
  • Semantic tokens: Maintain this canonical list: bg, bg-muted, bg-raised, text, text-muted, border, border-strong, accent, accent-hover, accent-active, accent-muted, accent-contrast, danger, warning, success, info, overlay, overlay-strong.
  • Palette guidance: Map semantic tokens to palette primitives defined in the CSS build pipeline; when a new primitive is needed, extend the palette once and remap consumers rather than hard-coding colors downstream.
  • Icon contrast rule: Icons and glyph-only buttons must meet a minimum 3:1 contrast ratio against their background; prefer accent-contrast or text tokens when in doubt.
  • Accessibility expectations: Default themes must satisfy WCAG 2.1 AA contrast ratios for text and controls, and preserve focus outlines sourced from the token set.
  • Prohibitions: Do not commit inline styles, <font> tags, or CSS variables whose values are literal colors; always reference the appropriate token variables.
  • Theme toggling: Implement new themes through token swaps only (e.g., toggling the data-theme attribute) and keep component logic agnostic to specific palette values. Document toggles alongside the README CSS build pipeline.

3. Magnet Handling — Do’s & Don’ts

  • Use the helpers every time: Always run inbound values through safeDecodeMagnet() before playback or normalization, and call normalizeAndAugmentMagnet() (see js/magnetUtils.js) to append ws= / xs= hints without mutating hashes.
  • Keep magnets raw: Persist the literal magnet:?xt=urn:btih:... string. Decode once, then feed that exact value to WebTorrent or storage. Never rely on re-encoding helpers that might alter casing.
  • Never call new URL() on magnets: URL constructors/URLSearchParams percent-encode the xt payload and can corrupt legacy hashes. Manipulate magnets with the helper utilities or string functions that respect the raw hash.
  • HTTP hints: Encourage HTTPS ws= web seeds (file roots only) and HTTPS xs= pointers to .torrent files so fallback peers warm quickly.
  • Tracker policy: Browser code must ship WSS trackers only. Pull the canonical list from js/constants.js; do not add UDP or plaintext HTTP trackers. (If you introduce a new tracker set or feature flag, update js/constants.js and mention it here.)
  • Validation reference: When in doubt about legacy infohash handling or normalization, consult the existing tests that cover legacy/bare infohash inputs and magnet normalization. If you extend helper behavior, extend those tests alongside the code.

4. Upload Modal Troubleshooting

  1. Required fields: Validation must ensure a title exists and that either a hosted URL or a magnet (or both) is supplied. Optional ws / xs inputs should be appended only when a magnet is present.
  2. Magnet parsing errors: When a user pastes encoded magnets, run safeDecodeMagnet() before normalizing. Show inline errors if decoding fails and confirm the raw magnet:?xt= string survives the round trip.
  3. Playback source promise: After submission, confirm the resulting feed item carries data-play-url and data-play-magnet attributes so playback utilities can attempt the configured source first.
  4. Mixed-content warnings: If the page is served over HTTPS, reject ws= or xs= hints that begin with http:// and show guidance to upgrade to HTTPS. Document the copy in README/UX strings so operators can align messaging.
  5. Telemetry hooks: Keep existing analytics/logging untouched unless instructed. If you add new modal states, annotate them clearly in code comments.
  6. Modal regressions: If validation or helper wiring breaks in Main, disable new feature flags and ship a revert PR immediately. Note the rollback steps in AGENTS.md for posterity.

4a. Cloudflare/S3 upload-path gotchas (audited 2026-06-16)

Invariants: js/services/r2Service.js and js/services/s3UploadService.js are near-identical copies — any fix to one must be mirrored in the other; storage keys must stay content-addressed (reuse computeStorageContentHash in js/r2.js for any new upload path); R2Service.deleteVideoStorage() stays best-effort, never throws, and only touches URLs under the owner's bucket base. Full writeup: docs/lessons/cloudflare-s3-upload-gotchas.md.

5. Manual QA Checklist

Run this script before shipping notable UI changes, especially around upload/playback flows. Automate when possible, but manual verification is required for releases.

  1. Smoke test the modal

    • Launch the site, open the Upload modal, and confirm required-field validation (title + one of URL/magnet).
    • Submit with URL only, magnet only, and both; ensure success states are clear.

  2. Hybrid playback

    • Publish a post with both URL and magnet. Load the card and verify the player fetches the configured preferred source first (check network tab or logs).
    • Temporarily block the primary source (e.g., devtools throttling or offline). Confirm playback seamlessly falls back to the secondary source.

  3. Magnet hygiene

    • Paste an encoded magnet; ensure safeDecodeMagnet() returns the raw value and the final string still includes the original xt hash.
    • Confirm normalizeAndAugmentMagnet() appends ws=/xs= hints without corrupting parameters.

  4. P2P hints

    • Verify magnets authored through the modal include HTTPS ws= and optional xs= values when provided.
    • Ensure tracker lists come from js/constants.js and remain WSS-only.

  5. Cross-browser sanity

    • Spot-check playback in Chromium and Firefox (desktop) to ensure no console errors related to CORS, Range requests, or tracker connections.

Document the run in PR descriptions so QA can cross-reference results.

6. Moderation Quickstart (bitvid)

The moderation and admin list tooling remain active in bitvid and must follow Nostr moderation policies and QA steps. Review docs/moderation/README.md and these key expectations:

  1. Report thresholds: Parse NIP-56 reports and compute trustedReportCount using only F1 followers.
  2. Default behaviors: Blur thumbnails at ≥3 F1 nudity reports and block autoplay at ≥2.
  3. Admin lists: Use NIP‑51 lists — 10000 for mute/downrank and 30000 for admin hard-hide/whitelist. Ensure these list IDs are correctly handled in the profile modal.
  4. Profile Modal Integration: Admin list toggles and moderation controls in ProfileModalController must trigger local state changes and reflect mutations visually without breaking cached profiles.
  5. User overrides: Add a “show anyway” option to any blurred or blocked content.
  6. Testing: Run all cases from docs/moderation/testing.md before shipping changes. Confirm blur, hide, and override flows function correctly across sessions.
  7. Telemetry and privacy: Log moderation decisions locally for debugging without leaking PII.
  8. QA demo: Record a ≤90 s screen demo showing blur/hide/override flows when submitting major moderation work.

7. Content Schema v3 & Playback Rules

  • Event payloads: New video notes serialize as version 3 with the JSON shape: { "version": 3, "title": string, "url"?: string, "magnet"?: string, "thumbnail"?: string, "description"?: string, "mode": "live"|"dev", "isPrivate": boolean, "deleted": boolean, "videoRootId": string }.
  • Validation: Every note must include a non-empty title plus at least one playable source (url or magnet). URL-only and magnet-only posts are both valid. Legacy v2 magnet notes stay readable.
  • Upload UX: The modal collects a hosted HTTPS URL and/or magnet. Enforce HTTPS for direct playback while keeping magnets optional when a URL is supplied.
  • Playback orchestration: bitvidApp.playVideoWithFallback probes and plays the HTTPS URL first, watches for stalls/errors, and then falls back to WebTorrent. When both sources exist, pass the hosted URL through to WebTorrent as a webseed hint.
  • Status messaging: Update modal copy to reflect whether playback is direct or via P2P so regressions surface quickly during QA.

8. Architectural Guide: bitvidApp vs. UI Controllers

To keep the codebase maintainable, we are moving complex UI logic from js/app.js into dedicated controller classes (e.g., ProfileModalController, UploadModal). This is an architectural direction for unstable—some controllers may not exist yet. When working on these components, follow these core principles:

Principle: Separation of Concerns

  • bitvidApp (js/app.js): The Orchestrator. Its main job is to manage application-wide state, wire up services, and orchestrate high-level actions. It should not directly manipulate the DOM elements of complex components like modals. Instead, it should instantiate their controllers and communicate with them through a clean API.

  • UI Controllers (e.g., js/ui/profileModalController.js): The Specialists. A controller is responsible for a specific piece of the UI. It manages its own DOM elements, handles user interactions within its scope, and uses services and callbacks provided by bitvidApp to perform actions.

State Management: The Source of Truth

The single source of truth for shared application state (like saved profiles, relay lists, and block lists) lives in modules like js/state/cache.js and services like relayManager.

  1. bitvidApp Writes State: bitvidApp is responsible for calling the functions that mutate this central state (e.g., setSavedProfiles, userBlocks.addBlock).

  2. bitvidApp Notifies the Controller: After updating the state, bitvidApp must call a method on the controller to tell it that its data is stale and it needs to re-render.

    • Example: When a user logs in, bitvidApp updates the active user and saved profiles, then calls this.profileModalController.handleAuthLogin(detail) so the modal can update its own display.

  3. Controller Reads State: The controller receives functions to read this state in its constructor (via the state and services objects). It should always use these functions (e.g., this.state.getSavedProfiles(), this.services.userBlocks.getBlocked()) to get the data it needs to render.

Execution Flow: Callbacks

When a user performs an action inside a modal that needs to affect the whole application (like switching a profile or updating relays), the flow is:

  1. User Action: A user clicks a "Switch Profile" button inside the modal.
  2. Controller Event Handler: The controller's internal event listener fires.
  3. Controller Invokes Callback: The controller calls a callback function passed to it by bitvidApp (e.g., this.callbacks.onRequestSwitchProfile({ pubkey })). It does not contain the logic for switching profiles itself.
  4. bitvidApp Executes the Logic: bitvidApp provides the implementation for that callback (e.g., handleProfileSwitchRequest). This method contains the actual logic to call the auth service, reload videos, and update the application state.

Checklist for Refactoring from bitvidApp to a Controller:

When moving a feature to a controller, follow these steps:

  1. Identify Logic: Find the methods in js/app.js that directly read from or write to the modal's DOM (e.g., old renderSavedProfiles or populateBlockedList logic).

  2. Create Controller Method: Implement the new logic inside the controller. Ensure it gets its data from this.services.* or this.state.*.

  3. Create bitvidApp Wrapper: In js/app.js, replace the old logic with a new, thin wrapper method. This wrapper should:

    • a. Update any central application state if needed.
    • b. Call the new method on the controller instance (e.g., this.profileModalController.populateBlockedList()).

  4. Wire Up Callbacks: If the controller needs to trigger an app-wide action, ensure a callback is defined in the controller's constructor and provided by bitvidApp during instantiation.

  5. Remove Old Code: Once fully replaced, delete the old DOM manipulation logic and direct event listeners from js/app.js. All interactions should be routed through the controller.

File Map (quick reference)

  • js/app.js — bitvidApp (orchestrator; owns high-level flows like playVideoWithFallback wiring)
  • js/services/playbackService.js — Lower-level playback stream handling
  • js/playbackUtils.js — Magnet/session helper utilities
  • js/magnetUtils.js — safeDecodeMagnet, normalizeAndAugmentMagnet (WSS trackers, ws=/xs= hints)
  • js/constants.js — Feature flags, canonical WSS tracker list

9. Nostr Addressing & NIP-33 (Lessons Learned)

  • Addressing Rule: NIP-33 events (e.g., videos, long-form content) are uniquely identified by the combination of kind, pubkey, and the d tag.
  • The d tag is King: When creating pointers for these events, always use the d tag as the identifier.
  • Avoid Logical IDs: Do not confuse logical identifiers (like videoRootId) with the network identifier (d tag). While they might often be the same, they are not guaranteed to be.
  • Hydration Pitfall: If you create a pointer using a logical ID that differs from the d tag, relays will fail to find the event during hydration because the index relies on the d tag.
  • Fix Pattern: When resolving a pointer from an event, prioritize checking tags.find(t => t[0] === 'd')?.[1] over any other ID field.

10. Direct Messages (NIP-04 vs NIP-44)

  • Dual Support: We must support both legacy NIP-04 and newer NIP-44 Direct Messages.
  • Reasoning: A significant portion of the user base still relies on NIP-04 DMs. While NIP-44 offers better security, we cannot drop NIP-04 support until adoption of the new standard is nearly universal.
  • Implementation: When working on DM features, ensure code paths exist for both encryption/decryption methods. Do not deprecate or remove NIP-04 handling logic. Other legacy methods can be considered for removal, but NIP-04 is critical.

11. Telemetry & Error Aggregation

  • Tooling: Use npm run telemetry:aggregate (which runs scripts/agent/telemetry-aggregator.mjs) to collect, sanitize, and aggregate errors from CI/test runs, smoke tests, and agent logs.
  • Privacy Policy: The tool automatically strips PII (IPs, emails, hex keys, bech32 keys) from logs. Do not manually inspect raw logs containing PII unless necessary for debugging in a secure environment; rely on the sanitized reports in ai/reports/.
  • Opt-in: Telemetry generation is opt-in. You must set the environment variable ENABLE_TELEMETRY=true when running the aggregator.
  • Reports: Aggregated reports are generated in ai/reports/telemetry-YYYYMMDD.md and artifacts/error-aggregates.json. These files provide prioritized issues for triage.

12. Multi-Agent Coordination

This project uses multiple AI agents (Claude Code, OpenAI Codex, Google Jules) working in parallel. Without coordination, agents will create conflicting PRs, duplicate work, and cause painful merge conflicts. Follow these rules to keep things clean.

Before Starting Work

  1. Check open PRs. Run the following command to list all open PRs: 
    curl -s "https://api.github.com/repos/PR0M3TH3AN/bitvid/pulls?state=open&per_page=100" | jq '{count: length, titles: [.[].title]}'
    If another agent already has a PR touching the same files or subsystem, do not create a competing PR. Instead, note the conflict and ask the maintainer how to proceed.
  2. Check this section for in-flight work. The maintainer may list active work areas below. Respect these reservations.
  3. Read KNOWN_ISSUES.md. Do not open PRs to fix issues already documented there unless explicitly asked.

Subsystem Boundaries

To minimize merge conflicts, treat these as independent work zones. An agent should avoid touching files outside its assigned zone in a single PR:

Zone Key Files Description
Nostr Core js/nostr/client.js, js/nostr/adapters/, js/nostrClientFacade.js Protocol client, signers, relay management
Event Schemas js/nostrEventSchemas.js, docs/nostr-event-schemas.md Event definitions and documentation
Playback js/services/playbackService.js, js/playbackUtils.js, js/magnetUtils.js Video streaming, magnet handling
UI Controllers js/ui/ Modal controllers, notification, components
State js/state/ Profile cache, application state
DMs js/ui/dm/, js/nostr/dm/ Direct messaging subsystem
Moderation js/moderation/, docs/moderation/ Content moderation, reports, admin lists
Build & CI .github/workflows/, scripts/, tailwind.config.cjs CI pipeline, lint scripts, build tooling
Styling css/tokens.css, css/tailwind.source.css Design tokens and Tailwind source
App Orchestrator js/app.js, js/app/ Main app wiring (high conflict risk — only one PR at a time)

js/app.js is the highest-risk file. It wires everything together. Never have two agent PRs modifying it simultaneously.

PR Discipline

  • One subsystem per PR. A PR that touches Nostr Core and UI Controllers is two PRs. Split them.
  • Small and mergeable. Target PRs that can be reviewed in one sitting. If a refactor spans 10+ files, break it into sequential PRs.
  • Merge before branching. Before starting new work on a subsystem, ensure all open PRs for that subsystem are merged or closed.
  • Rebase, don't stack. Long-lived branches diverge fast with multiple agents. Rebase onto unstable frequently.

Signaling Work to Other Agents

When you create a PR or start work, leave a clear signal:

  • PR title prefix: Use a descriptive prefix like [nostr-core], [playback], [ui], [ci] so the scope is visible at a glance.
  • PR description: Include a "Files Modified" section listing the key files touched, so other agents can quickly detect conflicts.

Task Coordination

The previous TORCH relay-locking protocol (and its torch/ tooling) has been removed. Coordinate via GitHub instead: before starting, check open PRs (gh pr list or the GitHub API) for overlap on the same subsystem, and use a descriptive PR title prefix so scope is visible. Don't run two agents on the same subsystem simultaneously, and keep js/app.js single-writer (see the Multi-Agent Coordination rules above).

Currently In-Flight Work

No reservations currently active.

13. Code Health Rules (CI-Enforced)

Three automated lint checks protect the codebase from common growth problems. These run in CI via npm run lint and will block your PR if violated.

File Size Limits

Script: npm run lint:file-size (scripts/check-file-size.mjs)

  • New .js files under js/ must stay under 1,000 lines.
  • Existing oversized files are grandfathered with a recorded line count. They must not grow beyond their recorded count + 50 lines.
  • If your change would push a grandfathered file over its limit, extract logic into a new module first, then make your change.

Decomposition strategy for agents:

  1. Identify a cohesive block of logic (e.g., a group of related methods, a subsection of a controller).
  2. Extract it into a new file in the same directory.
  3. Export the extracted functions/class.
  4. Import and re-wire from the original file.
  5. Run npm run lint:file-size to confirm the original file shrank.
  6. Update the grandfathered count if needed (the maintainer will handle this when merging).

innerHTML Baseline

Script: npm run lint:innerhtml (scripts/check-innerhtml.mjs)

  • Every file has a baseline innerHTML assignment count. Adding new innerHTML usage beyond the baseline fails the lint.
  • For new UI code, use safe DOM APIs instead: 
    // PREFERRED: auto-escaped, no XSS risk
element.textContent = userProvidedString;
const el = document.createElement("div");
el.className = "card";
parent.appendChild(el);

// ACCEPTABLE: innerHTML with escapeHtml for all user data
import { escapeHtml } from "../utils/domUtils.js";
container.innerHTML = `<span>${escapeHtml(title)}</span>`;

// BLOCKED: innerHTML with unescaped user data
container.innerHTML = `<span>${title}</span>`; // XSS vulnerability
  • If you must add innerHTML (e.g., loading a view template), update the baseline: node scripts/check-innerhtml.mjs --update and include the updated BASELINE in your PR.

Constants — No Duplication

  • Relay timing constants live in js/nostr/relayConstants.js — always import from there.
  • Before defining a new numeric constant, search for existing ones with the same value. If a match exists, import it.
  • Feature flags live in js/constants.js.
  • Cache policies live in js/nostr/cachePolicies.js.
  • Do not define the same constant in two files. If two modules need the same value, create a shared module and import from it.

14. Playwright Agent Testing Infrastructure

bitvid ships a test harness (window.__bitvidTest__, enabled via ?__test__=1), a mock relay (scripts/agent/simple-relay.mjs), and a reusable Playwright fixture (tests/e2e/helpers/bitvidTestFixture.ts) so agents can programmatically log in, seed relay data, and inspect app state — no browser extensions or real relays required. Full reference (harness API, mock relay, fixtures, data-testid selectors): docs/testing/agent-playwright-harness.md.

15. Agent Execution Protocol

This section defines the meta-workflow that every AI agent must follow when executing tasks in this codebase. It ensures safe, incremental changes, a clear decision trail, and continuity across context resets or agent hand-offs.

Mindset

  • Implement changes safely. Match existing conventions. Leave a clear trail of decisions.
  • Correctness > cleverness. Consistent conventions > personal preference. Explicit tradeoffs > "should work."
  • Prefer minimal, incremental changes over rewrites. Keep the build and tests green.
  • Do not invent files, APIs, libraries, or behaviors. If unsure, inspect the codebase first.

Scoping the Work

Before writing any code, establish clarity on three things:

  1. Primary goal — What is being built or fixed?
  2. Success criteria — How will we know it's done? (e.g., tests pass, feature visible in UI, lint clean)
  3. Non-goals — What is explicitly out of scope? (Prevents scope creep across agent sessions.)

External Notes (Persistent State Files)

Agents lose context across sessions. These files act as persistent memory that any agent (or human) can read to resume work or understand project status. Update them before any likely context reset.

Storage Rule: To prevent conflicts, these files are stored in dedicated directories with timestamps.

  • Read: Always look for the latest timestamped file in the directory (e.g., ls context/ | sort | tail -n 1).
  • Write: Never overwrite an existing file. Create a new file with a new timestamp (e.g., context/CONTEXT_YYYY-MM-DD_HH-MM-SS.md).
Directory File Pattern Purpose When to Update
context/ CONTEXT_<timestamp>.md Current goal, scope, assumptions, constraints, and "Definition of Done" At start; when scope changes
todo/ TODO_<timestamp>.md Checkbox task list with "Done" and "Blocked/Questions" sections After planning; after tasks
decisions/ DECISIONS_<timestamp>.md Key choices made, alternatives considered, and rationale When making tradeoffs
test_logs/ TEST_LOG_<timestamp>.md Commands run and their results (including failures) After every lint/test/build run

Rules for external notes:

  • These files are working documents. They should reflect the current state of work.
  • When starting a fresh task, clear stale content (by writing a fresh file) and re-scope.
  • Keep entries concise — bullet points and checklists, not prose.
  • decisions/ is especially important: when you make a tradeoff, write it down so the next agent doesn't undo it.

Work Loop

For each task item, repeat this cycle:

A) Locate

  • Identify the relevant code and existing patterns. Find the "right place" to make the change.
  • Write a short plan (3–7 bullets) before editing. If the task touches multiple subsystem zones (see Section 12), flag this — it may need to be split into separate PRs.

B) Implement

  • Make the smallest change that satisfies the requirement.
  • Keep edits cohesive. Avoid unrelated refactors in the same commit.
  • Follow all conventions already documented in this file (logging via logger, token-first styling, magnet safety, schema definitions in nostrEventSchemas.js, etc.).

C) Verify

  • Run the most relevant lint/test commands (see CLAUDE.md "Available Scripts" for the full list): 
    npm run lint              # Required
npm run test:unit:shard1  # Fast feedback; run full suite before PR
  • If there's no test coverage for your change, add focused tests or document in decisions/ why tests aren't feasible.
  • Log all commands and results in a new test_logs/TEST_LOG_<timestamp>.md file.

D) Update Notes

  • Check off completed items in the latest todo/ file (by creating a new one).
  • Record decisions in a new decisions/ file.
  • Log verification results in a new test_logs/ file.
  • Do this before any likely context/memory reset — if you're about to hit a token limit, finish a long session, or hand off to another agent, update notes first.

Context Recovery

If an agent starts a session and finds existing work-in-progress:

  1. Immediately read the latest files in context/, todo/, decisions/, and test_logs/.
  2. Resume from the next unchecked item in the latest todo/ file.
  3. Verify the current state by running git status, git log --oneline -10, and a quick lint/test pass.
  4. Do not re-do work that the latest test_logs/ shows already passed — unless the code has changed since.

Output Requirements

When finishing a task or handing off, provide:

  • Summary of what changed and why.
  • Files changed (list).
  • Verification results (from the latest test_logs/ file).
  • Follow-ups / risks / remaining TODOs.
  • If blocked: Ask specific questions and propose 1–2 options.

Quality Bar

Principle Meaning
Correctness over cleverness Simple, working code beats elegant abstractions
Conventions over preference Match what's already in the repo, even if you'd do it differently
Explicit tradeoffs Write down why in decisions/, don't leave the next agent guessing
Logged verification Every claim of "tests pass" must have a test_logs/ entry to back it up
Incremental over sweeping Small, reviewable changes that keep the build green at every step

16. Scenario-First Tests & Test Integrity (Dark Factory Standard)

Invariant: validation replaces code review in this repo — never weaken, delete, or rewrite a test merely to make it pass, and never mask flakiness with retries/sleeps/looser assertions. Expectation changes are allowed only via the Spec Correction Protocol with an accompanying Test Integrity Note. Read the full standard before adding or modifying any test: docs/testing/test-integrity.md.

17. NIP-07 Signer Reliability & Encrypted-List Loading (hard-won, 2026-06-16)

Invariant: the #1 cause of "DMs / hashtags / watch-history / lists won't load after login" is an unresponsive NIP-07 signer, not bitvid — diagnose with a raw window.nostr probe first, and do not regress the five resilience invariants (capped relay fan-out, NIP-07 circuit breaker, re-thrown transient decrypt errors, generous decrypt budgets, lazy modal population). Full writeup: docs/lessons/nip07-signer-reliability.md. See also docs/KNOWN_BUGS.md #0.

Repo memory

Curated agent memory lives in .agents/ (index .agents/MEMORY.md); read it first. Propose additions in .agents/proposals/; .agents/memory/ changes only via reviewed commits. Code/tests/config outrank memory.

Next

Read docs/nostr-event-schemas.md and js/nostrEventSchemas.js next; keep new nostr kinds centralized there and the doc up to date.

Agent fleet

This repo uses agent-fleet-kit. The agent-fleet MCP server (fleet-mcp) is a convenience. Editing backlog JSON and mailbox files directly is fine. Do not hand-edit the landing queue while a drain is merging, and do not push main.

  • area_info for the owner of a path.
  • backlog_claim with an item id marks it in-progress. Pick the item from PROGRAM.md. Without an id the tool does not choose one. Idle is a correct result.
  • mailbox_read at the start of a turn; mailbox_send to reach a peer. Mail is always delivered. The prompt says when to act.
  • commit to stage and commit in your area worktree. Codex workspace-write keeps .git read-only, so do not run git commit yourself. Pass message and the file paths (or all: true), and cwd or area when this server is not already in the worktree. submit: true also queues the tip. A doesNotOwn path warns and still commits. The same operation is fleet commit -m "..." --area <id> [--submit] -- <paths>.
  • landq_submit queues the committed tip and starts a drain. A dirty worktree warns; only HEAD is queued. Then go back to work. At your next natural checkpoint, land_status (or fleet landq status <id>) is a quick lookup, and its wait is optional. Never infer the lander or the land result from silence. States: queued, gating, landed with the main sha, rejected with the reason, or unknown. If no drain is alive, the answer says so and names fleet lander start.
  • A scarce machine is a prompt convention: say so in .task and wait if a peer has it. There is no slot lock.

Sessions, worktrees, and cron are the fleet CLI. Submit starts the drain; do not run a second merge by hand.

Reference in new issue View Git Blame Copy Permalink