Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/neXenio/bifrost-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/nexenio/bifrost-plugin/bifrost-candidates)<a href="https://agentmods.dev/commands/nexenio/bifrost-plugin/bifrost-candidates"><img src="https://agentmods.dev/badge/commands/nexenio/bifrost-plugin/bifrost-candidates/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/nexenio/bifrost-plugin/bifrost-candidates"><img src="https://agentmods.dev/badge/commands/nexenio/bifrost-plugin/bifrost-candidates.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00017 | $0.01223 |
| Opus 5 | $0.00009 | $0.00611 |
| Sonnet 5 | $0.00003 | $0.00245 |
| Haiku 4.5 | $0.00002 | $0.00122 |
Grade A, and why
bifrost-candidates scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/bifrost-candidates
Review the findings this project's sessions recorded, and decide which become shared team knowledge.
Candidates live in .bifrost/candidates.md in the project root (git-ignored). They are
written by the Stop hook's periodic memory check-in and are local and unreviewed —
nobody else can see them, and nothing reads them automatically. This command is that
reader.
Why this step exists
The shared corpus has no read permission model. memory_search does take filters —
wing, room, tier, agent_id, conversation_id, include_expired — but every one
of them is the caller's own retrieval narrowing, chosen per query and dropped just as
easily on the next one. Nothing a writer can set keeps an entry away from a reader. So
anything written to the corpus is recalled by every colleague immediately, as settled
team knowledge. Combined with correction machinery that exists but has never once been
used, a wrong entry is effectively permanent. Shared submission is therefore deliberate
and privacy-gated—never automatic retrieval. luca-memory v0.42 removed the collective
review workflow that once staged submissions for a vote, so a store now lands in the
corpus directly. Nothing downstream will catch a bad entry; this gate is the only one.
Procedure
1. Read the file. If .bifrost/candidates.md does not exist or holds only its
header, say so and stop — nothing has been recorded here yet.
2. Summarize. List each candidate with its date. Group near-duplicates; the same finding often gets recorded by more than one session.
3. Triage each one. For each candidate, judge:
- Still true? Check it against the current code. A finding recorded three weeks ago may already be stale, and stale is the failure mode that matters most here.
- Durable, or transient? A convention or root cause is worth keeping. "Fixed the auth test" is not.
- General, or session-specific? It must make sense to someone who was not there.
- Safe to share? No secrets, credentials, customer or personal data. Everything
promoted is recalled into colleagues' prompts and therefore sent to the model
provider — treat it as publishing to the whole company.
Run the classification check in
guidance/memory-classification.mdbefore promoting anything. First failing gate decides; when unsure, leave it in the spool.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 95 lines · 17 tokens per session scan A 3a96a978a85e
bifrost-candidates is a command published in the GitHub repository neXenio/bifrost-plugin (2 stars, last pushed 20d ago), licensed MIT. It adds 17 tokens to every session and 1,223 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
recall-save
Generate / overwrite .recall/context.md with Recall's local offline summarizer.
show
Show this project's saved Recall context (.recall/context.md).
ccc-context-budget
Visual context window budget analyzer — shows usage gauge, identifies bloat sources, recommends when to save session and compact.
ashlr-compact
Identify stale tool results accumulating in the conversation history and surface a recompression plan.
pin
Break checkpoint — status display, session log, optional MEMORY.md update.
atlas
Cross-project atlas with an adaptive render — ≤4 projects in scope get detail cards, more get one aligned digest line each, grouped by domain. Facets projects · sessions · memory · plans · specs · plugins · claude.md; a bare run renders all of them except plugins (opt-in, machine-global). --dir NAME|PATH anchors the…