recallloom

RecallLoom

RecallLoom is a portable context harness for session-based agents.

It provides a lightweight file model for project continuity across sessions without requiring heavy infrastructure.

The goal is not to remember everything. The goal is to keep the right project state durable, readable, and recoverable across sessions.

Package Scope

This file is the agent-facing entrypoint for the installable recallloom/ skill package.

Install and trigger this package through your host agent's normal skill discovery flow. RecallLoom itself does not require a custom host-specific launcher inside the package. The package may still ship optional native wrapper templates for supported hosts.

This installable package is intentionally kept lean. Human-facing repository landing pages and marketing docs may exist upstream, but they are not bundled into the installed skill directory.

In the source repository, README.md and README.en.md are concise public front doors, README.zh-CN.md is the compatibility entry, INDEX.md is the full map, and USAGE.md is the operator guide. Those files describe the same helper contract as this installed package entrypoint rather than defining a second logic set.

For package inventory, protocol details, and helper-script behavior, rely on the files that ship inside the package itself:

• managed-assets.json
• package-metadata.json
• references/file-contracts.md
• references/operation-playbooks.md
• references/package-support-policy.md
• references/protocol.md

Package Facts

<!-- RecallLoom metadata sync start: package-metadata -->

• package version: 0.4.3
• protocol version: 1.0
• supported protocol versions:
• 1.0

<!-- RecallLoom metadata sync end: package-metadata -->

Runtime Assumptions

<!-- RecallLoom metadata sync start: runtime-assumptions -->

• Python 3.10 or newer
• supported workspace languages:
• en
• zh-CN
• supported bridge targets:
• AGENTS.md
• CLAUDE.md
• GEMINI.md
• .github/copilot-instructions.md

<!-- RecallLoom metadata sync end: runtime-assumptions -->

Package Support Gate

RecallLoom package support is separate from project sidecar protocol compatibility.

• Helpers MUST perform the package-support check and MUST NOT write support state into project .recallloom/.
• If support is readonly_only, mutating helpers MUST block while diagnostic and read-only helpers MAY continue.
• If support is diagnostic_only, only diagnostic helpers SHOULD continue.
• If support is unknown_offline, diagnostic and read-only actions MAY continue, but mutating actions MUST block until support can be verified.
• Blocked actions MUST return the shared failure contract with blocked_reason: package_support_blocked and a package_support object. See references/package-support-policy.md.

Public Surface And Required Checks

• Public package and release surfaces MUST stay limited to files a user needs to install, understand, and operate the package.
• Public surfaces MUST NOT include copied project memory, generated runtime output, machine-local data, maintainer-only working files, or material that is not required by the installable package.
• Public CI and required checks MAY validate repository contents and metadata.
• Required-check wording MUST NOT present repository checks as proof of a user's local workspace state, host behavior, or sidecar trust status.

Non-Invasive Defaults And UX Gates

• Core install and daily use MUST NOT require or auto-install hooks, daemons, watchers, MCP/plugin enforcement, host adapters, telemetry/metrics, or remote payload transmission.
• Native command wrappers are opt-in convenience entrypoints over the same dispatcher, not a required enforcement layer.
• Ordinary docs/source/planning edits outside the managed sidecar stay silent allow or low-friction unless they affect provenance-sensitive RecallLoom state.
• Managed sidecar or provenance-impacting actions surface one of allow, warn, ask, or block in helper readiness output when provenance state is relevant.
• warn is for low-risk structural-only or readable legacy states and should stay brief; repeated same-session low-risk warnings should be cooldown-friendly.
• ask is for legacy review / repair import or reviewed imported baseline actions and requires explicit operator confirmation before higher-risk writes.
• block is non-waivable for forged markers, detected receipt/store inconsistency, direct state.json / config.json edits, privacy violations, and any state classified as inconsistent_or_tampered_evidence.
• Do not present remote services, host memory, plugins, MCP, hooks, or wrappers as authority for local helper evidence.
• Receipt-backed mutation is limited to dispatcher-issued managed-file writes, daily-log appends to the current latest cursor, and post-append summary sync. Archive apply and bridge apply remain preview-only until those surfaces gain their own receipt support.

Write Protocol Red Lines

• Managed sidecar writes MUST use helper scripts. Do not bypass them with blind file replacement, blind patching, or hand-built sidecar files.
• Daily-log writes MUST use append_daily_log_entry.py or dispatcher append. Do not handwrite daily-log-entry markers.
• Overwrite-style managed files MUST use revision-aware helper commits. Do not handwrite file-state markers.
• STORAGE_ROOT/state.json and STORAGE_ROOT/config.json MUST NOT be hand-edited during normal operation.
• Protocol 1.0 daily-log counters are file-local: entry-seq is 1..N within one daily log and canonical entry-id is entry-{entry_seq}. Do not treat either as globally unique.
• Keep state.json.daily_logs.entry_count as entry_count; it means the entry marker count in the latest active daily log, not a global cumulative count.
• If a helper write fails, diagnose, fix, retry, then surface the helper failure contract if it still cannot complete.
• Damaged-sidecar recovery MUST use the canonical recovery proposal/review/promotion helpers and validate_context.py; do not hand-edit managed markers, state.json, config.json, receipts, or helper-evidence stores.

When To Use It

Use RecallLoom when you need to:

• continue an existing project after a pause
• restore project context from maintained files
• maintain current-state project memory
• record meaningful milestone progress
• reduce context drift across sessions or tools

Typical triggers include:

• continue this project
• restore project context
• pick up where we left off
• rl-init
• update the project memory
• record today’s progress
• prepare a clean next-step handoff inside the maintained project files

First Attach Behavior

On first explicit invocation in a project, RecallLoom should not assume the workspace is already initialized.

The correct flow is:

1. detect whether a valid RecallLoom sidecar already exists
2. if it exists, continue normally without making initialization into extra ceremony
3. if it does not exist, explain that the project is not initialized yet and ask whether initialization should be performed
4. if the user explicitly confirms, or directly says rl-init, run the standard initialization action
5. if the environment cannot provide Python 3.10+, stop with a blocked runtime result instead of hand-building a sidecar

rl-init SHOULD mean: initialize the sidecar, validate the workspace, and return next recommended actions. Treat it as a stable high-level action name even when the host does not expose native slash commands.

Current Action Surface

For the current package line, the stable operator-facing wrapper targets are:

• rl-init
• rl-resume
• rl-status
• rl-validate

rl-init is the primary operator-friendly first-attach action name. The others are operator-facing stable action names that can be interpreted by the host agent or mapped into native custom commands when the host supports that surface. rl-bridge remains the canonical d