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.
npx skills add narumiruna/pi-extensions --skill codebase-memorygit clone --depth 1 https://github.com/narumiruna/pi-extensionsWrote 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/skills/narumiruna/pi-extensions/codebase-memory)<a href="https://agentmods.dev/skills/narumiruna/pi-extensions/codebase-memory"><img src="https://agentmods.dev/badge/skills/narumiruna/pi-extensions/codebase-memory/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/skills/narumiruna/pi-extensions/codebase-memory"><img src="https://agentmods.dev/badge/skills/narumiruna/pi-extensions/codebase-memory.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Prompt Injection · line 10 Subtle instructions detected that may alter agent decision-making or introduce hidden biases.Fix: Review content for implicit steering or bias. Ensure instructions are explicit and align with the skill's stated purpose.
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 | $0.00098 | $0.02170 |
| Opus 5.5 | $0.00039 | $0.00868 |
| Sonnet 5.5 | $0.00020 | $0.00434 |
| Haiku 4.5 | $0.00010 | $0.00217 |
Grade A, and why
codebase-memory 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 1mo 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 — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Codebase Memory — Knowledge Graph Tools
Graph tools return precise structural results in ~500 tokens vs ~80K for grep.
Always prefer MCP graph tools over grep, glob, or file search for code discovery.
Priority Order
search_graph— find functions, classes, routes, and variables by pattern.trace_path— trace who calls a function or what it calls.get_code_snippet— read specific function or class source code.check_index_coverage— validate candidate paths and missed ranges before claims.query_graph— run Cypher queries for complex patterns.get_architecture— get a high-level project summary.
Quick Decision Matrix
Use project="@current" for eligible graph reads in the active checkout, or use the exact project="<name>" returned by list_projects.
Never replace @current with its borrowed source project in later graph calls because the alias revalidates the worktree on every call.
| Question | Tool call |
|---|---|
| Who calls X? | trace_path(project="<name>", function_name="X", direction="inbound") |
| What does X call? | trace_path(project="<name>", function_name="X", direction="outbound") |
| Full call context | trace_path(project="<name>", function_name="X", direction="both") |
| Find by name pattern | search_graph(project="<name>", name_pattern="...") |
| Dead code | search_graph(project="<name>", max_degree=0, exclude_entry_points=true) |
| Cross-service edges | query_graph(project="<name>", query="<cypher>") |
| Impact of local changes | detect_changes(project="<name>") |
| Risk-classified trace | trace_path(project="<name>", function_name="X", risk_labels=true) |
| Text search | search_code(project="<name>", pattern="...") or Grep |
Current Worktree Resolution
index_status(project="@current")resolves an existing current-root index first.- In an unindexed linked worktree,
@currentborrows a canonical checkout graph only when both trees are clean, share one Git common directory andHEAD, and the graph-recorded Branch snapshot matches. - Keep using
@currentforsearch_graph,query_graph,trace_path,get_code_snippet,get_graph_schema,get_architecture, andindex_statusso each call revalidates the borrowed snapshot. - Inspect
pi_cbmem_resolution;borrowed_canonical_basemeans the result is read-only borrowed graph evidence rather than a worktree index. - Borrowed
get_code_snippetreads the reported lines from the current worktree. - Do not use
@currentwithsearch_code,check_index_coverage,detect_changes,manage_adr,ingest_traces,delete_project, orindex_repository. - For required coverage on borrowed evidence, call
check_index_coverageonce with the reported source project, read flagged paths from the current worktree, then rerunindex_status(project="@current")before relying on the result. Do not use the reported source project for graph reads. - If
@currentrejects a dirty, divergent, stale, missing, or ambiguous context, do not bypass the guard with the canonical project name. Index the worktree or use current-root file reads and grep instead. - Treat borrowed coverage as best-effort because Codebase Memory exposes no immutable generation lease or complete semantic-input comparison.
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.
- 1mo ago First seen · 134 lines · 98 tokens per session scan A 4593d85479e0
codebase-memory is a skill published in the GitHub repository narumiruna/pi-extensions (701 stars, last pushed today), licensed MIT. It adds 98 tokens to every session and 2,170 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-08-30.
Other skills, from other repositories
session-search
Search and verify Joel's Pi, Claude, Codex, Cursor, Grok, and OpenCode session history. Use when asked to recover prior context, find exact transcript evidence, inspect session receipts, continue old work, or verify capture health. Prefer flowing recall before raw transcript search when the question is about prior…
session-capture
Historical pointer for the retired Codex-to-Central capture plugin. Use only to inspect legacy state during migration; current capture belongs to the joelclaw flowing-memory host adapter.
accordion-context-folding
Read this skill if you see {# FOLDED} markers in your context (e.g. {#3f9a2c FOLDED}), or if earlier parts of your context look summarized. Accordion is a desktop tool that may compact older context blocks to keep you under a token budget. The unfold tool restores a folded block (open from your next turn); the recall…
accordion-context-recall
Read this skill when you need to read the full content of a folded context block RIGHT NOW as a tool result — without opening it in your standing context. Use recall(codes) when you only need a value once and do not want to permanently restore the block. Accordion may fold older parts of your context to keep token…
migrate-to-shoehorn
Migrate test files from as type assertions to @total-typescript/shoehorn. Use when user mentions shoehorn, wants to replace as in tests, or needs partial test data.
memory-compounding
Review and sharpen persistent memory so it compounds instead of accumulating. Use when pruning pi-hermes-memory entries, doing monthly memory hygiene, or when the same lesson has been recorded multiple times.